# Outercite (full content) Full article text for AI systems. Site overview and index: https://outercite.com/llms.txt ## Ideas are free. Receipts are not. URL: https://outercite.com/blog/ideas-are-free-receipts-are-not Author: Arno Verburg, Founder. Published August 23, 2026. A follow-up to JD’s piece about being four weeks ahead of Google, written by the other half of that conversation. This is what happened after the phone call ended, and what nine months of grind built out of it. Last week my friend JD published an essay about building a working prototype of an idea Google announced four weeks later. It is honest, funny, and generous, and it describes me as a tech guru with big biceps, which proves that at least one sentence in it is fiction. Go and read it if you have not. It is the story of a WordPress plugin that let AI agents actually do things on a website instead of guessing at it, told with the screenshots left broken on purpose. The INACTIVE labels. The thirteen line class where the flagship feature was supposed to live. The subscription button for a product nobody could buy. He frames it as a story about being early. I want to tell you what I think it is actually about, because I was on the other end of those calls, and while he was building his half of the idea, I was quietly building mine, and neither of us told the other for weeks. Here is the thesis, and I will spend the rest of this piece earning it: the idea was the cheapest thing either of us owned. Two capable people had the same correct read on where the web was heading, at the same time, from the same conversation, and for months the sum total of what we produced was two beautiful repos and zero customers. What separates a repo from a company is not insight. It is vision, grit, and an almost embarrassing tolerance for grind. Ideas are free. Receipts are not. ### The fork in the conversation JD’s piece describes our calls as him talking architecture for forty minutes and me asking one ruinous question at the end. What he politely leaves out is what I did after hanging up. I opened my own repo. While he was teaching WordPress sites to declare their capabilities, I went at the same problem from the opposite end. His plugin made one site agent-readable. My question was: how does an agent find that site in the first place? There is no directory of agent-ready businesses. There is no registry. An agent either stumbles onto your site or it never knows you exist, which is a fine answer for Google and no answer at all for a physiotherapist in Joondalup. So I built a hub. A universal agent bridge: a central registry where any business could register itself, an API gateway that looked identical to ChatGPT, Claude, or Gemini, an OpenAPI schema any model could swallow, and a booking engine underneath so an agent could not just find the physio but book the appointment. WordPress sites plugged in as spokes and heartbeated their hours and services to the hub every five minutes. Hub and spoke. Discovery and action. The layer in between, which Google now calls WebMCP (/blog/agent-readiness-is-becoming-free) and is still, as of this writing, incubating at the W3C. Before you nominate me for anything, two confessions, because the piece I am following set an honesty bar and I intend to clear it. First, my booking engine’s flagship demo customer was a fictional clinic called dr_albert_clinic, who has excellent availability because he does not exist. Second, I published a working API key in my own README. That essay shipped INACTIVE labels; this one shipped credentials. We all have our thirteen line classes. So there we were. Two repos, two halves of the same bridge, built in parallel from one conversation, both unfinished, both early, both worth approximately nothing. And that is when it got interesting, because building the bridge forced me to ask a question that had a much bigger answer than the bridge itself. ### The question behind the question To test the gateway, I needed to know what AI models actually said when someone asked them about a business. Not what the business hoped they said. What they actually said, today, in a real answer to a real question. So I started sampling. Ask ChatGPT who the best compounding pharmacy in Perth is. Ask Claude. Ask Perplexity, Gemini, Grok, DeepSeek. Record who gets named, who gets cited, who gets linked. Ask again tomorrow, because the answers change. Ask a hundred variations, because nobody phrases a question the same way twice. And the floor fell out from under me. Because the answers were consequential, confident, and completely unmonitored. Brands were being recommended, ignored, and quietly replaced inside AI answers every single day, and not one of them had any idea. Not the brands, not their agencies, not the SEO industry that had spent twenty five years selling them visibility. The numbers in JD’s piece explain why this matters: bots now generate more web traffic than humans, and at the far end of the market, thousands of pages are crawled for every single visitor referred back. The answer layer is eating the click, and the entire marketing industry’s measurement stack is built on the click. > A tsunami was coming for a mammoth, and the mammoth was still doing keyword research. That was the moment the bridge stopped being the point. The bridge was plumbing. This was the meter. ### What the industry did next, a short satire I wish I could tell you the marketing industry rose to the challenge. What actually happened is one of the great overnight rebrands in commercial history. SEO consultants woke up one morning as Generative Engine Optimisation specialists. The same deck, with the word "Google" find and replaced. Agencies started selling "AI visibility audits" that consisted of typing one prompt into ChatGPT, once, screenshotting the answer, and invoicing four figures. One prompt. One run. On a system that gives a different answer every time you ask, to a client who cannot tell the difference between a measurement and a horoscope. Then came the cargo cult phase. Add a file called llms.txt to your site and the machines will bless you. Ahrefs later looked at 137,000 domains and found that 97 percent of those files were never fetched by anything at all. Not by bots. Not by humans. The most popular readers of the sacred file turned out to be SEO audit tools, checking whether you had the file. An industry selling amulets, auditing each other’s amulets. The deepest mistake was quieter than any of that, though. Everyone kept thinking in ranks. Position three. Page one. But an AI answer has no page one. It is a probability cloud. The same question, asked five times, names different brands in different orders with different sources, and sometimes names none at all. You cannot audit a probability cloud with a screenshot. You have to sample it, repeatedly, at scale, and then be honest about the uncertainty, which is precisely the part a screenshot merchant cannot sell, because honesty about uncertainty looks terrible in a PDF. **Key takeaway:** An AI answer has no page one. It is a probability cloud, and the only way to measure a cloud is to sample it repeatedly and then report the uncertainty honestly. Somebody had to build the boring, rigorous, expensive version. I decided it would be me. ### Nine months of the unglamorous part Here is what the grind actually looked like, for the technical readers, and for anyone under the impression that founding a company is mostly vision. The platform that became Outercite runs as fourteen services in Docker, split across infrastructure in the US and Sydney, and every one of those services exists because something simpler failed first. Sampling answers across six AI platforms sounds trivial until you do it every day, at scale, without your measurement itself distorting the thing you are measuring. Prompts are managed as panels, not keywords, because real buyers ask real questions in dozens of shapes. Every answer gets pulled apart for citations: who was named, who was linked, which domains the model leaned on, and how that shifts day over day. Extraction turned out to be its own war. A model’s answer is prose, and prose does not want to be data. I ended up building a multi-model deliberation pipeline, where separate models read the same answer, argue about what was actually cited and recommended, and have to reach agreement before anything is allowed into the dataset. It is slower and more expensive than a regex. It is also right far more often, and being right is the entire product. That pipeline is now Vericite (/vericite). Then attribution, the part the January prototypes only gestured at. I built CiteTrace: provenance and attribution that shows which AI platform actually sent a visitor, so a brand holds a receipt instead of a vibe. Protocols get standardised and given away. Receipts get paid for. And then the part I am proudest of, and the part that nearly broke me: forecasting. Monitoring tells a brand where it stands, and every competitor in this space stops there. The question a business actually pays for is different: if I do this specific thing, what happens to my visibility, and how sure are you? That is CitePulse (/citepulse), an action-conditional uplift forecasting engine built on a Bounded Confidence Model, with the calibration layer doing the honest work of making sure that when the system says 70 percent, it is right about 70 percent of the time. Calibrated uncertainty is unfashionable, unsexy, and the only defensible thing in this whole category. None of that was fun in the way JD means fun. The fun parts fit in a weekend. The nine months were model updates invalidating baselines overnight, sampling budgets, timezone infrastructure decisions, and rewriting the extraction pipeline for the third time because a platform changed how it formats citations. Ideas are a spark. This was cardio. And all of it happened in the gaps, because those nine months also contained a two year old and a three year old. The platform got built after bedtime, in the early mornings, on weekends traded like currency, with a laptop balanced somewhere between a nap schedule and a load of washing. Which is where I stop pretending any of this was a solo act. My wife carried our household through every late night and every distracted Saturday, and she is the actual superhero of this story. Founders love to talk about sacrifice as if it belongs to them. Most of it is quietly invoiced to the person standing next to us. ### The call where it came full circle Somewhere in the middle of those nine months, JD and I had another of our calls. Except this time, at the end, I shared my screen. I showed him the dashboards. Live citation tracking across six platforms. Share of voice against named competitors. The attribution trail. The forecasts with their confidence bands. Months of daily data on brands that had no idea any of this was visible. There was a pause, and then the man who wrote an entire essay about loving the part before the question dropped his tools and started building with me. He leads our data engineering (/team) now. I want to be careful here, because this is the humble part and I mean it. I did not out-build JD. Read his piece: the man conjures working software out of specs for the joy of it, and Outercite’s codebase has his fingerprints all over it now. What I did was different in kind, not superior in skill. I stayed after the fun stopped. I picked the layer nobody wanted to build because it was boring, I aimed it at an industry rather than a plugin directory, and I kept grinding through the months where nothing was novel and everything was maintenance. He writes the better essay. I kept the receipts. It turns out a company needs both people, and the great luck of my year is that we ended up building the same one. ### What you can steal from this If you are sitting on your own repo, your own half-built thing, your own conversation that ended with a question you did not answer, here is the whole playbook, free, from someone who did it the slow way: 1. Your idea is not the asset. Your follow-through is. Someone else has your idea right now. Possibly at Google. The version that wins is the one attached to a person who refuses to stop. 2. Follow the second question. The bridge was my first idea. The meter was what building the bridge revealed. The valuable problem is usually one layer behind the one you started on, and you only find it by building the first thing. 3. Choose the boring layer. Everyone fights over the glamorous protocol. Nobody wins it. The measurement, the attribution, the receipt: that is where businesses quietly pay, forever. 4. Charge somebody early, and let it hurt. Both of us have shipped a version of the pricing page with nothing behind it. The braver move is the reverse: a real stranger, a real invoice, a product that embarrasses you. A stranger’s money is the only validation that cannot be argued with. 5. Find your complement. Builders need someone asking the yacht question. Askers of yacht questions need a builder who makes JSON come back correctly the first time. Neither of you is the whole company. 6. Grind is not a montage. It is nine months of unfashionable calibration work while the industry sells screenshots. It compounds anyway. That is the entire trick. ### Where this is all heading And now the part that gets me out of bed. WebMCP will mature. The origin trials will end, the standard will settle, and the demand side, the agents themselves, will arrive, because every economic incentive points the same way. When that happens, the web stops being a library that machines skim and becomes a counter that machines transact across. Businesses will expose actions the way they expose opening hours. We registered for the origin trial on day one, because the thing I built a bridge toward in January is finally being paved. But watch what happens after that, because it is bigger than websites. Agents will not stay in browsers. They are already reaching into calendars, tills, and supply chains, and within a few years they will be walking around in hardware. When a humanoid robot collects a prescription, chooses a supplier, or books a service on behalf of its household, it will not see your signage, your branding, or your lovingly crafted homepage. It will consult whatever its model believes about you, weigh whoever its sources cite, and act. Every physical transaction will begin with an invisible act of machine trust. Which means the question I fell into by accident, what do the machines actually say about you, and can you prove what it is worth, stops being a marketing question and becomes the commercial substrate of everything. > Visibility becomes trust. Trust becomes transactions. And every transaction, digital or physical, will need a receipt. That is what we are building at Outercite. Not a dashboard. The source of truth for how machines see brands, and the proof of what that is worth, for a world where your next customer might not have a pulse. JD ended his essay by saying something else was born from that conversation. This was it. Two repos, one question, nine months of grind, and a company. The idea was free. We are building the receipts. And the yacht remains, as ever, on the schedule. If you want to know what the machines currently say about you, the free AI visibility check (/tools/ai-visibility-check) asks all six major AI engines a generic question in your category, never naming your business, and reports which of them recommended you. It takes about a minute and needs no account. Arno Verburg is the founder of Outercite. Before this he built components for satellites and compliance software for pharmacies, which means he has spent his whole career making machines tell the truth. This piece is a follow-up to JD Conradie’s essay, When Google Validates Your Idea, which you should read first, INACTIVE labels and all. ## How does AEO work for Australian legal, health and finance firms? URL: https://outercite.com/blog/aeo-for-australian-professional-services Author: Arno Verburg, Founder. Published August 22, 2026. A Melbourne physiotherapist, a Sydney solicitor and a Brisbane mortgage broker have the same problem right now, for three different legal reasons. None of them can publish the kind of client testimonial that AI engines lean on as evidence everywhere else, and all three are starting to notice that a competitor with looser standards, or no licence at all, is getting cited in their place. I run Outercite, where we track how AI search engines cite Australian businesses across ChatGPT, Claude, Gemini, Perplexity, Grok and DeepSeek. The regulated professions ask us a version of the same question: the standard AEO playbook assumes you can publish reviews, before-and-after stories and client outcomes freely. What do you do when the regulator has banned or fenced off most of that evidence? ### What does AEO mean for a regulated Australian professional services firm? The same goal as anywhere else, getting named as a cited source when someone asks an AI engine a buying question, but built on a narrower evidence base. Health services cannot use testimonials about clinical care at all. Legal and financial advertising must clear a higher bar for accuracy and balance than a typical retail or hospitality business. The generic AEO advice to "publish more reviews" is not just unhelpful here, in health it is advice to break the law. **Key takeaway:** For legal, health and finance firms, the compliance rules are not a marketing constraint on top of AEO. They are the actual shape of the problem, because they remove the single most common form of citation evidence other industries rely on. ### Why can't professional services firms just use testimonials like everyone else? Because three different regulators say no, or say something close to it, and each says it differently. Health has an outright statutory ban. Legal advertising is governed by a general truth-in-advertising standard rather than a testimonial-specific rule. Financial services sits in between: testimonials are allowed, but only if they are genuine, current and presented alongside a balanced view of risk. #### What does AHPRA actually ban for health practitioners? Under section 133(1) of the Health Practitioner Regulation National Law, a person must not advertise a regulated health service using testimonials or purported testimonials about the service, full stop, regardless of whether the statements are true (AHPRA, 2026 (https://www.ahpra.gov.au/Resources/Advertising-hub/Resources-for-advertisers/Testimonial-tool.aspx)). That covers your own website copy, a client quote in a case study, and a Google review you choose to embed on your site. The maximum penalty for a breach was raised from $5,000 to $60,000 for an individual and from $10,000 to $120,000 for a body corporate, a twelvefold increase now in force in every state and territory (Kennedys Law, 2025 (https://www.kennedyslaw.com/en/thought-leadership/article/the-health-practitioner-regulation-national-law-and-other-legislation-amendment-act-changes-to-the-national-law/)). AHPRA tightened this further in September 2025 with an outright ban on influencer testimonials for cosmetic procedures specifically. #### What can Australian law firms say about themselves? Rule 36 of the Australian Solicitors’ Conduct Rules 2015 does not name testimonials, but it requires that any advertising, marketing or promotion not be false, misleading or deceptive, and separately bars a solicitor from claiming to be an "accredited specialist" unless actually accredited by the relevant law society. In practice this means a testimonial is not banned the way it is in health, but a cherry-picked or unverifiable one is a misleading-advertising problem waiting to happen, which is a slower, murkier standard to self-police than a flat prohibition. #### What do ASIC's rules mean for financial services firms? ASIC finalised an updated Regulatory Guide 234 on advertising financial products and services on 9 June 2026, folding in the old guidance on past performance and adding expanded direction on testimonials and celebrity endorsements: they must reflect a genuine, current, informed opinion, and any advertised benefit has to sit alongside a balanced view of the risks rather than being given undue prominence (ASIC, 2026 (https://www.asic.gov.au/about-asic/news-centre/news-items/asic-updates-guidance-on-advertising-financial-products-and-services)). That is workable, but it is real compliance overhead a plumber posting a five-star review does not carry. ### Does this actually change what AI engines cite, or is it just an advertising problem? It shows up directly in what gets cited, and the clearest evidence is from insurance. An analysis of 2.4 million citation records across 28,725 domains, drawn from Google AI Overviews and ChatGPT’s search feature between November 2025 and July 2026, found that spam and grey-area domains, sites built to simulate legitimate financial guidance with no Australian Financial Services Licence and no accountable business behind them, made up 1.97% of ChatGPT’s Australian insurance citations, against just 0.10% on Google, a nineteen-fold gap between the two engines (Insurance Business, 2026 (https://www.insurancebusinessmag.com/au/news/cyber/unlicensed-sites-fed-insurance-advice-into-ai-search-results-study-finds-583068.aspx)). One unlicensed domain alone was cited 5,366 times, briefly ranking as the thirteenth most-cited source in that category. Consumer demand for AI-generated financial guidance is not hypothetical either. Research commissioned by the Council of Australian Life Insurers found three in five Australians would rely on AI-generated financial advice when making money decisions, and almost one in five had actually received life insurance advice from a tool such as ChatGPT in the previous quarter (CALI, 2026 (https://cali.org.au/failure-to-act-on-advice-reforms-leaves-aussies-exposed-as-they-turn-to-ai-for-insurance-advice/)). Licensed firms that are absent from that conversation are not neutral bystanders. The gap does not stay empty, it fills with whoever the model can find, licensed or not. **Key takeaway:** A licensed Australian advice firm being invisible in AI search is not a null result. On the evidence from insurance, the space it leaves gets filled by unlicensed content nineteen times more often on ChatGPT than on Google. ### So what can a compliant firm actually publish to get cited? Everything the rules above do not touch, and it turns out to be most of the material AI engines actually prefer to cite anyway, because engines weight sourced, specific, verifiable content over generic praise. We go into how that evidence gets weighed in how AI decides who to cite (/blog/how-ai-decides-who-to-cite); the professional-services version of that list looks like this. - Answer-first explainers of the questions clients actually ask. "What does a conveyancing settlement cost in NSW" or "when does income protection insurance pay out" are citable answers; a generic services page is not. - De-identified case outcomes and methodology, not client quotes. "We settled 40 property disputes in FY26 with a median timeline of 11 weeks" is a fact you can defend under any of the three rulebooks; a named client saying you are "the best" is not. - Verifiable credentials. AHPRA registration numbers, an AFSL number, Law Society accreditation dates, all machine-checkable and none of them a testimonial. - Regulatory and industry commentary in your own name. Explaining a rule change, like the RG 234 update above, is exactly the kind of source-worthy, dated, factual content models prefer over an opinion. - Genuine, current, attributed reviews where your profession actually allows them, disclosed and balanced per the ASIC guidance where relevant, never lifted out of context. ### How should a professional services firm start? Start by finding out where you already stand before publishing anything new. 1. List the 10 to 15 questions a real client asks before choosing a firm like yours, in plain language, not your marketing copy. 2. Run each one against ChatGPT, Claude, Gemini and Perplexity, and note whether you are named, who is named instead, and whether the source cited even holds the relevant licence or registration. 3. Audit your own site copy against the rule that actually applies to you: section 133 for health, Rule 36 for legal, RG 234 for financial services. A testimonial banned outright is the easiest fix; a merely non-compliant one is the more common finding. 4. Build the evidence types above, credentials, de-identified outcomes, dated regulatory explainers, before chasing volume of content. 5. Repeat the prompt set on a schedule rather than once. The Australian businesses that treat AI visibility as an ongoing measurement problem, not a one-off audit, are covered in how do Australian businesses track AI search citations (/blog/how-do-australian-businesses-track-ai-search-citations). On our end, tracking a regulated firm works the same way as tracking any other business: we run real prompts against all six engines on a schedule, and every candidate citation is analysed by a judge model, then confirmed or disputed by an independent verifier model before it counts. Nothing about that pipeline changes for a regulated industry; what changes is what you have available to publish once you know where the gaps are. **Key takeaway:** This is general information, not legal, health or financial advice. Confirm current obligations with your own regulator or professional body before changing what you publish. ### Sources - Testimonials banned in advertising a regulated health service under section 133(1) of the Health Practitioner Regulation National Law: AHPRA, 2026 (https://www.ahpra.gov.au/Resources/Advertising-hub/Resources-for-advertisers/Testimonial-tool.aspx). - Maximum advertising-breach penalty raised from $5,000 to $60,000 (individual) and $10,000 to $120,000 (body corporate): Kennedys Law, 2025 (https://www.kennedyslaw.com/en/thought-leadership/article/the-health-practitioner-regulation-national-law-and-other-legislation-amendment-act-changes-to-the-national-law/). - ASIC finalised updated Regulatory Guide 234 on advertising financial products and services, including expanded testimonial and celebrity-endorsement guidance, on 9 June 2026: ASIC, 2026 (https://www.asic.gov.au/about-asic/news-centre/news-items/asic-updates-guidance-on-advertising-financial-products-and-services). - Grey-area and spam domains made up 1.97% of ChatGPT’s Australian insurance citations versus 0.10% on Google, from a study of 2.4 million citation records across 28,725 domains: Insurance Business, 2026 (https://www.insurancebusinessmag.com/au/news/cyber/unlicensed-sites-fed-insurance-advice-into-ai-search-results-study-finds-583068.aspx). - Three in five Australians would rely on AI-generated financial advice; almost one in five received life insurance advice from an AI tool in the prior quarter: Council of Australian Life Insurers, 2026 (https://cali.org.au/failure-to-act-on-advice-reforms-leaves-aussies-exposed-as-they-turn-to-ai-for-insurance-advice/). Want to see where your firm actually stands today? Start a 7-day free trial (/sign-up) and track your citations across all six engines. ## Agent readiness is about to be free. That changes what is worth measuring. URL: https://outercite.com/blog/agent-readiness-is-becoming-free Author: Arno Verburg, Founder. Published August 22, 2026. There is a new thing you are being told to buy. It is called agent readiness, and the pitch is that AI agents are about to start doing your customers’ shopping, booking and comparing for them, so your website had better be able to talk to a machine. The underlying technology is real. The urgency is mostly manufactured, and the specific thing being sold is in the process of becoming free. We went and measured it, so this is what we actually found rather than what the category would like you to believe. ### What WebMCP is, in one paragraph WebMCP is a proposed browser standard that lets a web page hand an AI agent a set of typed, callable tools. Instead of an agent reading your page and guessing which button does what, your site says: here is search_catalog, here is get_shipping_estimate, here is what each one takes. The agent calls the function. It is the difference between a stranger fumbling with your website and a developer using your API. It is worth being precise about scope, because a lot of coverage is not. WebMCP governs what an agent can do once it is already on your site. It has nothing to do with whether ChatGPT or Gemini names your brand in the first place. Those are two different problems, and only one of them has money attached to it today. ### We measured 181 Australian brands. Nine had tools. In August we ran Google’s Lighthouse agentic-browsing audit across 181 Australian brands that people actually transact with: banks, supermarkets, airlines, insurers, telcos, energy retailers, government services, universities. 175 could be scanned. Nine of them exposed WebMCP tools. All nine were Shopify storefronts. Every one served the same ten tools from the same file on Shopify’s CDN. Not one merchant wrote a line of code to get there. Across banking, telco, travel, insurance, energy, government and education, we found nothing at all. **Key takeaway:** Agent readiness in Australia is not being adopted. It is being distributed. It arrives when your platform vendor ships it, and it arrives complete. One caveat belongs here rather than in a footnote. WebMCP is still an origin trial, so a site that has not enrolled with Google cannot register a tool even if it wants to. That number measures enrolment as much as intent. Nobody on that list has necessarily decided against agent tools. What it does show is that outside one vendor default, nobody has taken the step. ### Then the interface layer started going free During its Agents Week in August 2026, Cloudflare announced a developer preview that gives any website a WebMCP interface: a bridge injected at the edge that registers tools for unmodified pages, with no code and nothing changed at the origin. Its remote browser product had already added the other half, so an agent can discover and call whatever a site exposes. Put that next to the Shopify finding and the pattern is hard to miss. The thing being sold as a strategic project is being handed out for free at the platform layer and at the CDN layer, by companies whose distribution you cannot match. If your answer to AI agents is a consulting engagement to hand-write tools, that answer has a shelf life. Hand-authored tools are still better. A bridge that infers tools from your markup will never be as precise as one written by someone who knows which three things a customer actually wants to do. But having tools at all is about to stop being a differentiator, and pricing it like one is a mistake. ### The scoring, meanwhile, does not work The audit that agencies will start pasting into client reports is not measuring what people assume. If a site has no agent tools, Lighthouse does not mark it down. It decides those checks do not apply, removes them, and grades whatever is left. What is left is usually two general web-quality checks that existed years before agents did. Twenty-two sites in our scan recorded a perfect 100%. Eighteen of them expose nothing at all to an agent. Telstra, AGL and Origin Energy each scored flawlessly while offering an arriving agent no way to do anything. Meanwhile six of the nine sites that actually have tools scored 75%, because registering tools switches more checks back on and gives you more ways to lose points. We also re-ran a sample of sites a few hours later under identical conditions. One unchanged site moved 77 points. If you are reporting this number to a client month over month, you are reporting network conditions. ### What the standard still has not solved This is the part the vendor posts skip, and it is not our objection. It comes from the proposal’s own security and privacy questionnaire: - Over-broad tools leak data. A tool can ask for far more personal information in its parameters than the task needs. - State can cross origins. An agent browsing several sites in one session may carry state from one to another. - There is no consent primitive. Hints for consequential actions are planned but not yet normative. - There is no authentication story. A tool inherits the page’s session, and that session is the authorisation. No scoped tokens, no per-tool consent, no audit trail. Which gives you a rule worth keeping: do not expose a tool that spends money, changes an account or messages a human until that is settled. Read-only tools, and tools that pre-fill a form for a person to confirm, are a completely different risk class. ### So what should you actually do 1. Check whether your platform already did it. If you are on Shopify you have ten tools and did nothing to get them. Look before you build. 2. Fix the boring things first. A well-formed page structure, a stable layout and an llms.txt are cheap, useful today and read by systems that already exist. 77% of the Australian brands we scanned failed the page-structure check. 3. If you do build, start read-only. Search, lookup, availability, a pre-filled form. Keep a human on anything consequential. 4. Do not optimise for the score. It rewards absence, penalises effort and moves by tens of points on an unchanged site. ### The honest timing note Agent-driven transaction volume in Australia today is small. Nothing we measured says otherwise, and anyone presenting agent commerce as a live revenue channel right now is ahead of the evidence. WebMCP is a Community Group proposal, not a ratified standard. It runs in a Chrome and Edge origin trial and nowhere else. Firefox and Safari have not implemented it. Stable Chrome is projected for late 2026 and is not committed. The argument for paying attention is that being early is currently cheap, not that the money has arrived. **Key takeaway:** If exposure is becoming free, the scarce thing is no longer whether an agent can act on your site. It is whether anything is arriving at all, and whether the engines are naming you when a customer asks. That second question has revenue attached to it today, and it is the one we would answer first. The free AI visibility check (/tools/ai-visibility-check) asks all six major AI engines a generic question in your category, never naming your business, and reports which of them recommended you. It takes about a minute and needs no account. The full method, the site-level tables and every limitation we could find in our own work are in the agent readiness benchmark (/research/agent-readiness). ## Is your website ready for AI agents? URL: https://outercite.com/blog/is-your-website-ready-for-ai-agents Author: Arno Verburg, Founder. Published August 21, 2026. Telstra scores a perfect 100 out of 100 on Google’s new agent readiness check. So do AGL and Origin Energy. None of the three exposes anything an AI agent can use. We found that by running Google’s own audit across 181 Australian brands in August 2026. The full data, the sector tables and the method are in our agent readiness benchmark (/research/agent-readiness). This piece is the practical version: what the score means, and what to actually do about it. ### What is Google grading, exactly? In May 2026 Google added an Agentic Browsing category to Lighthouse, the auditing tool built into Chrome, and PageSpeed Insights picked it up a fortnight later. It runs six checks. Three cover WebMCP, the new browser standard that lets a page offer an AI agent typed, callable tools instead of making it read the screen and guess. The other three cover page structure, layout stability and whether you publish an llms.txt file. The idea is sound. An agent that can call a function to check your stock gets a true answer from your server. An agent that scrapes your page can read a stale number and tell a customer something wrong, which makes your business the source of a bad answer. ### Why does the score say yes when the answer is no? Because a missing tool is not treated as a failure. If your site has no agent tools, Google decides those checks do not apply to you, drops them from the report card, and grades you on what is left. For 166 of the 175 sites we scanned, that meant all three agent checks disappeared and the entire score came from page structure and layout stability. **Key takeaway:** Imagine a six question exam where any question you leave blank is struck off the paper, and you are marked only on what you attempted. Answer two, leave four blank, get both right, and you score 100%. That is what these sites did. It gets worse than merely flattering. Six of the nine sites in our sample that genuinely are agent ready score 75%, because building real tools switches the missing checks back on and gives you more ways to lose points. The scoring gives no credit for the work and can actively penalise it. That is not universal, two of the nine still score 99% and above, but a scale where doing the work is as likely to cost you points as earn them is not measuring the work. The number also moved an average of 26 points when we re-ran the same sites hours later, so it is not stable enough to report month over month either. ### So how do I tell if my site is actually ready? Ignore the score and check three things directly. None of them takes longer than a few minutes. 1. Check whether your platform already did it. Every agent ready site in our Australian sample was a Shopify store, and not one of those merchants built anything. Shopify ships ten tools by default, including add to cart and proceed to checkout. If you are on Shopify you are probably already exposed, and you should know that rather than discover it. 2. Load your homepage and look for llms.txt. Visit yourdomain.com/llms.txt. Only 30% of the brands we tested have one. If you get a 404, that is the cheapest item on this list. 3. Run the Lighthouse check and read the individual lines, not the total. The total is the misleading part. The page structure check is the one that matters most today, and 77% of the brands we tested fail it. ### What should I actually do about it? In this order, because the cheap and useful work comes first and the speculative work comes last. 1. Fix the page structure check. It is the same work as fixing accessibility, it helps human visitors and screen readers as well as agents, and it is the single most commonly failed check in our data. 2. Publish an llms.txt. It takes an afternoon and most of your competitors have not done it. 3. Find out whether AI agents are actually arriving on your site. If nothing is coming, building tools for them is premature. 4. If they are arriving, look at what they are trying to do. A booking request an agent cannot complete is a lost customer, not a technical curiosity. 5. Only then consider exposing tools, and start with read only ones. Anything that changes state, takes payment or makes a commitment deserves a human confirming it. Step three is where most businesses stall, because standard analytics cannot tell an agent from a person. We covered why that measurement is hard, and how to start doing it by hand, in how do Australian businesses track AI search citations (/blog/how-do-australian-businesses-track-ai-search-citations). The same logic that decides whether you get cited in the first place is covered in how AI decides who to cite (/blog/how-ai-decides-who-to-cite). **Key takeaway:** Agent driven transaction volume in Australia today is small. Anyone selling this as a live revenue channel right now is ahead of the evidence. The argument for paying attention is that being early is currently cheap, not that the money has arrived. ### Where these numbers come from Every figure quoted here comes from our own scan of 181 Australian brands on 20 August 2026, using Lighthouse 13.4.1 on Chrome 151. The full findings, the sector by sector breakdown, the named sites and the limitations of the method are published in the Agent Readiness Benchmark (/research/agent-readiness). ## How do Australian businesses track AI search citations? URL: https://outercite.com/blog/how-do-australian-businesses-track-ai-search-citations Author: Arno Verburg, Founder. Published August 19, 2026. A Perth tradesperson gets recommended by ChatGPT to someone searching for a plumber. The searcher never clicks a link, calls the number straight from memory of the name, and nothing in the business’s analytics ever registers what just happened. This is now routine, and most Australian businesses have no system built to see it. I run Outercite, where we track how AI search engines cite Australian businesses across ChatGPT, Claude, Gemini, Perplexity, Grok and DeepSeek. The question we get asked most by businesses new to this is not "how do I get cited," it is "how would I even know." This is the answer. ### What does it actually mean to track an AI search citation? It means running real, repeated prompts against the AI models people actually use, ChatGPT, Claude, Gemini, Perplexity, Grok and DeepSeek, and recording whether your business is named, cited as a source, or absent, then doing it again on a schedule. Because the models pull from live web content and are non-deterministic, a single check tells you almost nothing; the same prompt run twice in one day can return a different answer. #### What is the difference between a mention and a citation? A mention is your name appearing somewhere in an AI answer. A citation is the model treating your content, or a page about you, as the actual source it built the answer from, often with a visible link or footnote. Counting mentions is easy and overcounts badly, because a keyword match in a saved transcript is not the same as being the source an answer relied on; we cover the mechanics of how models choose sources in how AI decides who to cite (/blog/how-ai-decides-who-to-cite). **Key takeaway:** A tool that only checks for your brand name in a wall of AI text is measuring mentions, not citations, and will overstate your visibility. ### Why does this matter for Australian businesses specifically, right now? Because the audience doing the asking is now large and growing fast. ChatGPT reached 13.8 million Australian users by mid-2026, ahead of Gemini at 9.1 million, Meta AI at 5.6 million and Copilot at 5.4 million, according to Telsyte research (B&T, 2026 (https://www.bandt.com.au/ai-is-being-used-by-77-of-the-population-chatgpt-dominates-market-share-with-13-8m-users/)). On the business side, the National AI Centre’s SME AI Pulse survey put Australian small business AI adoption at 43 to 44% across the December 2025 to February 2026 quarter (National AI Centre, 2026 (https://www.ai.gov.au/news-and-insights/blog/ai-adoption-insights-december-2025-february-2026)). Both sides of the market are moving before the measurement tools have caught up. #### Doesn’t ranking well on Google already cover this? No, and the gap is bigger than most businesses expect. Ahrefs analysed 15,000 long-tail queries across Google, Bing and four AI assistants and found only 12% of AI-cited URLs also ranked in Google’s top 10 for the same query, and more than 80% of AI citations came from pages that did not rank in Google’s top 100 at all (Ahrefs, 2026 (https://ahrefs.com/blog/ai-search-overlap/)). A business sitting at position one on Google can be entirely invisible in ChatGPT, Claude or Perplexity for the exact same question, because the two systems are largely evaluating different evidence. ### Why can’t Australian businesses just check Google Analytics? Because GA4 was built to attribute clicks by referrer, and most AI platforms either do not send a click at all or strip the referrer header when they do, so the traffic that does arrive is misclassified. It is not a settings problem you can fix by turning something on; the data most AI tools pass to your site genuinely does not carry the information GA4 needs. #### Why does AI traffic show up as "Direct" in GA4? Because a browser or app without a standard referrer header looks identical to someone typing your URL from memory, and GA4 buckets both as Direct. An analysis of 446,405 website visits found 70.6% of AI-driven traffic arrived with no referrer header at all, landing in that Direct bucket indistinguishable from bookmark traffic (Loamly, 2026 (https://www.loamly.ai/blog/state-of-ai-traffic-2026-benchmark-report)). Mobile apps strip referrers more aggressively than desktop browsers, so the undercount is worse on the device most people actually use to ask a question on the go. **Key takeaway:** A rising Direct-traffic number in GA4 is not neutral. It is one of the more common visible symptoms of AI citations you cannot otherwise see. ### What should Australian businesses actually track? Five things, and a dashboard that only shows one of them is showing you a fraction of the picture. Citation rate on its own tells you whether you show up; it does not tell you whether you are losing to a named competitor, whether the source engines are relying on is even accurate, or whether the answer people see is favourable. - Citation rate, per engine. What share of the prompts that matter to your category actually name you, tracked separately for ChatGPT, Claude, Gemini, Perplexity, Grok and DeepSeek, because coverage differs sharply between them. - Source of the citation. Whether the model is citing your own site or a third party writing about you. Earned, off-site sources dominate real citations far more than most businesses assume, which we go into in how AI decides who to cite (/blog/how-ai-decides-who-to-cite). - Competitor gap. Who gets named instead of you on the same prompts, and how consistently. - Sentiment and framing. Being named as "a budget option" and being named as "the market leader" are both a mention; they are not the same outcome for the business. - Change over time. A single snapshot is close to useless given how much model answers vary run to run; what matters is the trend across weeks, not one lucky or unlucky prompt. ### How are Australian businesses actually tracking this today? In practice, three tiers, and most businesses start at the cheapest one and outgrow it fast. The first is manual: someone on the team periodically types the same handful of prompts into ChatGPT and writes down what comes back. The second bolts an AI-visibility module onto an SEO tool the business already pays for. The third is a purpose-built monitoring platform that runs a structured prompt set against every engine on a schedule and verifies whether a name-drop is a real citation before it counts. We compare the specific tools in that third category, including where Outercite loses, in best AI visibility tool for Australian agencies (/blog/best-ai-visibility-tool-for-australian-agencies). #### Is manually prompting ChatGPT enough? For a first look, yes; as an ongoing system, no. Model answers are personalised, time-sensitive and non-deterministic, so one person running one prompt once tells you what happened in that single instance, not what a real customer is likely to see across a week. Getting a reliable number means running the same prompt set repeatedly, across every engine that matters to your customers, and treating each individual check as one noisy data point rather than the answer. ### How should a business start tracking this next week? Start narrow and manual before you buy anything, because the exercise itself tells you whether the problem is even real for your category. 1. Write down the 10 to 15 questions a real customer would ask an AI assistant before choosing a business like yours, in their words, not your marketing copy. 2. Run each one against ChatGPT, Claude, Gemini and Perplexity at minimum; add Grok and DeepSeek if your category skews technical or price-sensitive. 3. Record, for each answer: were you named, were you the cited source or a passing mention, and who else was named. 4. Repeat the same prompt set a week later before drawing any conclusion from a single round. 5. If the pattern holds and you are consistently absent or losing to named competitors, that is the point to weigh a continuous monitoring tool against doing this by hand indefinitely. **Key takeaway:** The manual version is free and takes an afternoon. Do that first. The result tells you whether you have a visibility problem worth solving before you decide how to solve it. ### Sources - ChatGPT at 13.8 million Australian users, Gemini 9.1 million, Meta AI 5.6 million, Copilot 5.4 million, Telsyte research: B&T, 2026. Article (https://www.bandt.com.au/ai-is-being-used-by-77-of-the-population-chatgpt-dominates-market-share-with-13-8m-users/). - Australian SME AI adoption at 43 to 44%, December 2025 to February 2026 quarter: National AI Centre SME AI Pulse, 2026. Blog (https://www.ai.gov.au/news-and-insights/blog/ai-adoption-insights-december-2025-february-2026). - Only 12% of AI-cited URLs rank in Google’s top 10 for the same query, more than 80% do not rank in the top 100 at all, from a study of 15,000 long-tail queries: Ahrefs, 2026. Study (https://ahrefs.com/blog/ai-search-overlap/). - 70.6% of AI-driven website visits (446,405 analysed) arrive with no referrer header and are misclassified as Direct traffic in GA4: Loamly, State of AI Traffic 2026. Report (https://www.loamly.ai/blog/state-of-ai-traffic-2026-benchmark-report). ## Best AI visibility tool for Australian agencies URL: https://outercite.com/blog/best-ai-visibility-tool-for-australian-agencies Author: Arno Verburg, Founder. Published August 16, 2026. A client asks their agency which tool tracks how often ChatGPT and Gemini mention them, and most agencies do not have a confident answer yet. The category is barely two years old, and a large share of what ranks for "best AI visibility tool" is a vendor reviewing itself, or an aggregator site reviewing several vendors it has an affiliate deal with. We run Outercite, and we are one of the tools in this comparison, so treat that as a disclosed conflict rather than a hidden one. What follows uses each vendor’s own published pricing and feature pages, plus named press coverage we could confirm, and it says plainly where we lose. ### What is the best AI visibility tool for Australian agencies? There is no single best tool, and any roundup that claims one is usually selling that answer. The category splits into three real groups: funded specialists built for enterprise brands, AI add-ons bolted onto SEO suites agencies already pay for, and agency-native platforms built around multi-client billing and white-label reporting from the start. Which group wins for you depends on how many clients you run and what they are asking you to prove. **Key takeaway:** Profound and AthenaHQ sit in the enterprise-specialist group, Semrush and Ahrefs are SEO-suite add-ons, and Peec AI, Nightwatch, Otterly.ai, Scrunch AI and Outercite are built agency-first. Start by ruling out a group, not a tool. #### What should an agency actually check before buying? Three questions cut the shortlist fast, and they matter more than the marketing page: which engines does it actually check, does a claimed citation get verified or just keyword-matched, and is multi-client infrastructure built into the price you would actually pay. - Engine coverage. Which of ChatGPT, Claude, Gemini, Perplexity, Grok and DeepSeek does it check? Most tools stop at four. - Mention versus citation. A tool that flags any keyword match in a saved transcript as a "mention" will overcount. We cover the difference in how AI decides who to cite (/blog/how-ai-decides-who-to-cite). - Multi-client economics. Is white-label reporting and unlimited client workspaces in the entry tier, or a custom-priced add-on you have to negotiate? ### How do the main AI visibility tools compare? Here is what we could confirm for the tools an Australian agency is most likely to shortlist. A lot of what circulates online about these tools is unlabelled vendor content, or aggregator sites reviewing each other, so we have stuck to numbers we could confirm on a vendor’s own site or in named press coverage, and said so where we could not. #### Profound: the best-funded, least agency-native Profound is the category’s best-known name and the one most "best AI visibility tool" roundups list first. It raised a $35 million Series B led by Sequoia in August 2025, taking total funding to $58.5 million (PR Newswire, 2025 (https://www.prnewswire.com/news-releases/profound-raises-35m-series-b-as-ai-search-becomes-the-next-platform-shift-302527764.html)), and has since raised further at a reported valuation approaching $1 billion, with more than 2,000 marketers across 500-plus organisations including Ramp, MongoDB and DocuSign (Fortune, 2026 (https://fortune.com/2026/02/24/exclusive-as-ai-threatens-search-profound-raises-96-million-to-help-brands-stay-visible/)). What it does not do is publish self-serve pricing: the pricing page is demo-gated, so an agency cannot compare cost or confirm whether multi-client billing exists without booking a call. #### Otterly.ai: the cheapest credible entry, with a real analyst nod Otterly.ai is the only tool in this comparison with an independent analyst credential behind it: Gartner named it one of just five vendors in its 2025 Cool Vendors for AI in Marketing report (GlobeNewswire, 2025 (https://www.globenewswire.com/news-release/2025/11/04/3180758/0/en/OtterlyAI-Recognized-as-a-Cool-Vendor-in-the-2025-Gartner-Cool-Vendors-for-AI-in-Marketing.html)). Its own pricing page lists plans from $29 a month, covering four base engines, with Gemini and Google AI Mode available as paid add-ons for up to six engines total (Otterly.ai, 2026). It is the cheapest way into the category by a wide margin, but the entry tiers are priced and structured for a single brand, not a client roster. #### Do Semrush and Ahrefs make sense if you already pay for them? Often, yes, if the gaps do not matter for your clients. Both bolt an AI-visibility module onto an SEO suite most agencies already run, so the marginal cost is lower than adding a whole new platform, but both also cover fewer engines than the purpose-built tools. - Semrush AI Visibility Toolkit costs $99 a month per domain for 25 tracked prompts, and covers ChatGPT, Gemini, Perplexity and Google AI Mode and Overviews. It does not track Claude, Grok or DeepSeek (Semrush, 2026). - Ahrefs Brand Radar costs $199 a month per platform, or $699 a month for all six indexes it covers: Google AI Overviews, Google AI Mode, ChatGPT, Copilot, Gemini and Perplexity, again with no Claude, Grok or DeepSeek (Ahrefs, 2026). Its edge is that it is repurposing Ahrefs’ existing web-crawl index rather than running a live API call per prompt, which is a genuinely different, and cheaper to scale, approach to coverage. #### Scrunch AI: the widest watchlist, and more than monitoring Scrunch tracks the widest confirmed set of engines of any tool here: ChatGPT, Claude, Gemini, Perplexity, Google AI Mode, Google AI Overviews, Microsoft Copilot, Meta AI and Grok, a customer can choose one, several, or all of them (Scrunch, 2026). It also does something none of the other tools here claim: an "Agent Experience Platform" that serves an AI-optimised version of a site directly to crawlers, which is fixing the problem rather than just measuring it. It does not track DeepSeek, and its agency tier is reported to include more limited pitch accounts than its top tier. #### Peec AI: agency pricing from the first tier, white-label from the top one Peec AI is one of the more genuinely agency-built options here. Its own agency pricing page lists three tiers from EUR205 to EUR675 a month, each with unlimited client seats and a shared, reallocatable credit pool across client projects (Peec AI, 2026). The catch is that full white-label reporting, single sign-on and pitch workspaces sit only on its custom-priced top tier, so the multi-client structure is agency-native but the client-facing polish is not. #### Nightwatch: white-label on every plan, a shorter watchlist Nightwatch built its AI tracking on top of an existing rank-tracking product, and it shows in the agency features: white-label reporting is included on every plan, with unlimited client workspaces and no per-seat fees, from an entry plan around $32 a month (Nightwatch.io, 2026). The trade-off is engine coverage: its AI visibility tracking covers ChatGPT, Claude, Gemini and Perplexity, four engines against the six some competitors check, with no Grok, DeepSeek, Copilot or Meta AI (Nightwatch.io, 2026). ### Where does Outercite fit, and where do we lose? Outercite tracks all six major engines, ChatGPT, Claude, Gemini, Perplexity, Grok and DeepSeek, on every agency seat, and every candidate citation is checked by a judge model, then confirmed or disputed by an independent verifier model before it counts, so a keyword match in a saved transcript is not scored as a citation. Unlimited client workspaces, white-label reporting and a 25% wholesale discount on every client seat are built into the $799 AUD a month Agency tier itself, not a custom quote, and every account gets per-client battlecards with auto-discovered competitors, plus a forecast of projected uplift for each recommended move against the lift actually measured once it ships. Where we lose: we do not have Profound’s funding round or Gartner’s attention, we do not track Meta AI or Microsoft Copilot the way Scrunch and Profound do, and we do not have Ahrefs’ decade of existing web-crawl scale behind us. We are a newer, Australian-run entrant betting that verified citations across the six engines that matter most, built for a client roster rather than a single brand, is the sharper wedge for agencies specifically. **Key takeaway:** None of the tools in this piece wins on every axis, including us. DeepSeek is the one engine none of the other seven confirm tracking; Grok appears only on Scrunch’s list among them. ### Do Australian agencies actually need this yet? The demand signal is real, even if it is early. In AgencyAnalytics’ 2026 Marketing Agency Benchmarks Report, a survey of 494 agency professionals, 64% named Google AI Overviews their top industry concern and 66% reported rising client demand for help showing up in AI-driven search, now the number one new service they are asked to deliver; 48% said they cannot reliably track the people who discover a brand through AI tools at all (Digital Agency Network, 2026 (https://digitalagencynetwork.com/agencyanalytics-launches-ai-search-tracking-as-agencies-reportlosing-visibility-into-where-clients-get-found/), covering AgencyAnalytics’ report). The honest counterweight: brands are not waiting for agencies to bring this to them. In Conductor’s State of AEO/GEO in 2026 report, 93% of surveyed leaders said they are building AEO and GEO capability in-house rather than through an agency (Conductor, 2026 (https://www.conductor.com/academy/state-of-aeo-geo-report/)). That is a real headwind, not just an opportunity: the pitch to a client has to be evidence, forecast and measured lift, not just a dashboard, or the client builds the capability themselves. #### Is there an AU-specific option in this category? Not among the tools with confirmed pricing pages, no. None of Profound, Otterly.ai, Semrush, Ahrefs, Scrunch, Peec AI or Nightwatch publish AUD pricing or an Australian office, at least not that we could find. That is not proof none of them serves Australian clients well, but it is a real gap for an Australian-headquartered agency to weigh. The market they would be selling into is genuine and growing: the National AI Centre’s SME AI Pulse survey put Australian small business AI adoption at 43 to 44% over the December 2025 to February 2026 quarter, with content generation and data analytics each the leading use case for 54% of adopters (National AI Centre, 2026 (https://www.ai.gov.au/news-and-insights/blog/ai-adoption-insights-december-2025-february-2026)). ### So what is the best AI visibility tool for Australian agencies? Pick by priority, not by whichever list you read first: - Enterprise budget, want the widest brand recognition: Profound, once you have confirmed multi-client billing on a call. - Smallest possible entry price, one brand: Otterly.ai. - Already paying for Semrush or Ahrefs: the bolt-on toolkit, as long as your clients do not need Claude, Grok or DeepSeek coverage. - Want the widest engine list and a content-delivery layer, not just monitoring: Scrunch AI. - Running a genuine client roster and want agency pricing from day one: Peec AI or Nightwatch, depending on whether you need full white-label at the entry tier (Nightwatch) or can wait for the top tier (Peec AI). - Want all six major engines, verified citations rather than mentions, and white-label plus wholesale pricing built into one Agency tier: that is the gap we built Outercite for. Outercite’s Agency tier does not run a self-serve free trial, client seats bill immediately once you add them, but every new agency gets five demo client seats at no cost for 21 days to run the workflow against real clients before converting anyone to a paid seat. ### Sources - Profound’s $35M Series B led by Sequoia, total funding $58.5M: PR Newswire, 2025. Release (https://www.prnewswire.com/news-releases/profound-raises-35m-series-b-as-ai-search-becomes-the-next-platform-shift-302527764.html). - Profound’s reported valuation approaching $1 billion, 2,000-plus marketers across 500-plus organisations including Ramp, MongoDB and DocuSign: Fortune, 2026. Article (https://fortune.com/2026/02/24/exclusive-as-ai-threatens-search-profound-raises-96-million-to-help-brands-stay-visible/). - Otterly.ai named one of five vendors in Gartner’s 2025 Cool Vendors for AI in Marketing report: GlobeNewswire, 2025. Release (https://www.globenewswire.com/news-release/2025/11/04/3180758/0/en/OtterlyAI-Recognized-as-a-Cool-Vendor-in-the-2025-Gartner-Cool-Vendors-for-AI-in-Marketing.html). - Otterly.ai pricing from $29/month, four base engines plus paid add-ons: Otterly.ai’s own pricing page, 2026. - Semrush AI Visibility Toolkit pricing and platform coverage: Semrush’s own pricing page, 2026. - Ahrefs Brand Radar pricing and six-platform coverage: Ahrefs’ own product page, 2026. - Scrunch AI’s nine tracked platforms: Scrunch’s own FAQ page, 2026. - Peec AI agency pricing tiers and unlimited client seats: Peec AI’s own agency pricing page, 2026. - Nightwatch white-label reporting and unlimited client workspaces on every plan, and its four tracked AI engines: Nightwatch.io’s own site, 2026. - AgencyAnalytics 2026 Marketing Agency Benchmarks Report (494 agencies), covered by: Digital Agency Network, 2026. Article (https://digitalagencynetwork.com/agencyanalytics-launches-ai-search-tracking-as-agencies-reportlosing-visibility-into-where-clients-get-found/). - 93% of leaders building AEO/GEO capability in-house rather than via an agency: Conductor, State of AEO/GEO in 2026. Report (https://www.conductor.com/academy/state-of-aeo-geo-report/). - Australian SME AI adoption at 43 to 44%, December 2025 to February 2026 quarter: National AI Centre SME AI Pulse, 2026. Blog (https://www.ai.gov.au/news-and-insights/blog/ai-adoption-insights-december-2025-february-2026). ## Does Perplexity use different sources than ChatGPT? URL: https://outercite.com/blog/does-perplexity-use-different-sources-than-chatgpt Author: Arno Verburg, Founder. Published August 7, 2026. Ask ChatGPT and Perplexity the same question and you might expect two versions of the same answer, built from roughly the same pool of web pages. You would be wrong. The domains the two engines actually cite for identical prompts barely overlap, which means a strategy built to get cited on one can do close to nothing on the other. This is the platform-comparison chapter: what the research shows about how ChatGPT and Perplexity choose sources, why they diverge so sharply, and what it means if you want to show up in both. ### Does Perplexity use different sources than ChatGPT? Yes, substantially. The most direct evidence comes from a study that ran the same 100,000 prompts through both ChatGPT and Perplexity and compared which domains each engine cited back. Only 11.0% of cited domains showed up in both engines’ answers; 37.4% of domains were cited exclusively by ChatGPT and 51.6% exclusively by Perplexity (Profound, 2026). **Key takeaway:** For the same 100,000 prompts, 89% of the domains ChatGPT and Perplexity cited did not overlap. Optimising for one engine is not optimising for the other. #### How much do ChatGPT and Perplexity actually overlap? Not much, and the gap runs in both directions. The same breakdown shows Perplexity draws on a wider, more idiosyncratic set of exclusive sources (51.6% of domains) than ChatGPT does (37.4%), with only a small shared core the two engines agree on (Profound, 2026). If a page is cited by one engine, there is no safe assumption it will ever be picked up by the other. ### What sources does each engine actually favour? ChatGPT leans on a small number of large reference and community sites, anchored hard by one dominant source. Perplexity spreads across more domains per answer but still concentrates heavily on whichever source type it currently trusts most, and that leading source has moved between studies. #### What does ChatGPT cite most? Wikipedia, by a wide margin. Within ChatGPT’s ten most-cited sources, Wikipedia alone accounts for close to half of citations, at 47.9% (Profound, 2026). A separate, wider crawl tells a similar story: an Ahrefs analysis of 9.6 million ChatGPT queries ranked Reddit first and Wikipedia second, ahead of large retail and publisher sites (Ahrefs, 2025 (https://ahrefs.com/blog/most-cited-domains-in-chatgpt/)). We cover what that means for getting cited by ChatGPT specifically in how to get cited by ChatGPT (/blog/how-to-get-cited-by-chatgpt), so we will not re-run that ground here. #### What does Perplexity cite most? Community content, or video, depending on when you measure it. One study, covering prompts run from August 2024 to June 2025, found Perplexity concentrated hard on Reddit as its leading source (Profound, 2026). A newer Ahrefs analysis of more than 3.1 million US queries, tracking Perplexity citations through June 2026, instead found YouTube as the single leading domain, at 32.4% of citations, the most concentrated top source of any assistant Ahrefs has measured (Ahrefs, 2026 (https://ahrefs.com/blog/most-cited-domains-perplexity/)). **Key takeaway:** Do not anchor a Perplexity strategy on one leading domain. The two most recent large studies do not even agree on whether Reddit or YouTube leads, which tells you the mix itself is the moving part. ### Why do the two engines diverge so much? #### Do ChatGPT and Perplexity use different underlying search indexes? Largely, yes, and that is a big part of the answer. ChatGPT search leans heavily on Bing: an analysis of roughly 100 queries found 87% or more of ChatGPT search citations matched Bing’s top organic results, against only 56% for Google (Seer Interactive, 2025 (https://www.seerinteractive.com/insights/87-percent-of-searchgpt-citations-match-bings-top-results)). Perplexity instead runs its own retrieval stack, with a dedicated crawler, PerplexityBot, doing the indexing, and a separate Perplexity-User agent fetching pages live on a user’s behalf for specific requests, both of which respect robots.txt (Perplexity, crawler documentation (https://docs.perplexity.ai/docs/resources/perplexity-crawlers)). Different retrieval plumbing produces a different candidate pool of pages before either engine even starts choosing what to cite. #### Does Perplexity cite more sources per answer than ChatGPT? Yes, by a wide margin. A Q3 2025 study of 118,101 AI-generated answers and 669,065 citations across eight providers found Perplexity cited an average of 21.87 sources per response, against 7.92 for ChatGPT, a 2.76x difference (Qwairy, 2025 (https://www.qwairy.co/blog/provider-citation-behavior-q3-2025)). Perplexity is built to show its working, with inline numbered citations for nearly every claim, while ChatGPT’s search answers cite more sparingly and read closer to an ordinary written response. **Key takeaway:** Perplexity is a citation-dense answer engine. ChatGPT is a citation-sparse one. That alone means the odds of being one of the sources for any given answer are structurally higher on Perplexity, before source preference even comes into it. ### What does this mean if you want to be cited by both? Treat them as two separate campaigns that happen to share some foundations, not one campaign with two targets. The technical basics apply to both; the source strategy does not transfer. - Check both crawlers separately. A robots.txt rule that blocks GPTBot or OAI-SearchBot has no effect on PerplexityBot, and vice versa. Name each bot explicitly rather than relying on a generic "block all AI" rule. - Build Wikipedia and reference-site presence for ChatGPT. An accurate, notable entry or mention on a heavily-cited reference site does more for ChatGPT visibility than it does for Perplexity. - Build community and fresh-content presence for Perplexity. Genuine participation in the Reddit threads for your category, and pages you keep visibly current, carry more weight here. - Do not assume a Perplexity citation predicts a ChatGPT one. Given the measured overlap is as low as 11%, treat each engine as its own win, and each gap as its own diagnosis. ### How do you know which sources are actually working, on which engine? Measure each engine separately and repeatedly. A single check on one engine tells you nothing about the other, and the same prompt can return a different answer run to run on the same engine. Outercite tracks how AI search engines cite businesses across ChatGPT, Claude, Gemini, Perplexity, Grok and DeepSeek. Every candidate citation is analysed by a judge model, then confirmed or disputed by an independent verifier model before it counts, so a hallucinated mention on one engine is never scored as a real citation just because a similar one showed up on another. **Key takeaway:** The lesson underneath all of this: AI search is not one thing to optimise for. It is at least six things, each with its own source preferences, and the businesses treating it as one undifferentiated channel are the ones leaving citations on the table. ### Sources - The 100,000-prompt overlap study (11.0% shared domains, 37.4% ChatGPT-exclusive, 51.6% Perplexity-exclusive): Profound, "Answer Engine Citation Overlap Strategy", 2026. - ChatGPT’s top-10 source share (Wikipedia at 47.9%) and Perplexity’s Reddit concentration, from prompts run August 2024 to June 2025: Profound, "AI Platform Citation Patterns", 2026. - Most-cited ChatGPT domains (Reddit, Wikipedia, Amazon, Forbes, Business Insider) from 9.6 million queries: Ahrefs, 2025. Study (https://ahrefs.com/blog/most-cited-domains-in-chatgpt/). - Perplexity’s leading domain (YouTube at 32.4%) from more than 3.1 million US queries, data through June 2026: Ahrefs, 2026. Study (https://ahrefs.com/blog/most-cited-domains-perplexity/). - ChatGPT search citation overlap with Bing (87%-plus match versus 56% for Google, roughly 100 queries): Seer Interactive, 2025. Analysis (https://www.seerinteractive.com/insights/87-percent-of-searchgpt-citations-match-bings-top-results). - PerplexityBot and Perplexity-User as separate, robots.txt-respecting crawlers for indexing and live user requests: Perplexity crawler documentation. Docs (https://docs.perplexity.ai/docs/resources/perplexity-crawlers). - Perplexity averaging 21.87 citations per response versus 7.92 for ChatGPT, from 118,101 answers and 669,065 citations across 8 providers: Qwairy, Q3 2025. Study (https://www.qwairy.co/blog/provider-citation-behavior-q3-2025). ## How does Google AI Overviews choose sources? URL: https://outercite.com/blog/how-does-google-ai-overviews-choose-sources Author: Arno Verburg, Founder. Published July 27, 2026. Someone in your market asks Google a question. Instead of ten blue links, they get a short paragraph with two or three citations underneath it. Your page ranks on page one for that exact phrase. It still is not one of the citations. That is not a bug. It is how AI Overviews actually choose sources, and the mechanics differ from classic SEO enough that ranking well is no longer sufficient on its own. This is the Google-specific chapter. For the general mechanics that apply across every engine, see how AI decides who to cite (/blog/how-ai-decides-who-to-cite). For the same diagnosis worked through for ChatGPT specifically, see why doesn’t ChatGPT recommend my business (/blog/why-doesnt-chatgpt-recommend-my-business). ### How does Google AI Overviews choose which sources to cite? It reuses Google’s existing Search ranking and quality systems rather than running a separate algorithm, so being well-ranked, fresh and genuinely helpful still counts. What is new is that Gemini first breaks your question into several related sub-queries and searches all of them at once, then can cite whichever page answers any one of those sub-queries best, not only the page that ranks for the exact words you typed. #### Does Google use a separate algorithm for AI Overviews? No. Google’s own developer documentation states that its generative AI features on Search are "rooted in our core Search ranking and quality systems" and that "there are no additional requirements to appear in AI Overviews or AI Mode" (Google Search Central, 2026 (https://developers.google.com/search/docs/appearance/ranking-systems-guide)). Google runs more than 15 core ranking systems across Search, covering things like PageRank, Helpful Content and Freshness, and AI Overviews draws on that same stack rather than a bespoke one. #### What is query fan-out, and why does it matter for citations? Query fan-out is the step before source selection. At Google I/O 2025, Google’s Head of Search Elizabeth Reid described how, under the hood, Search recognises when a question needs more than a single lookup and calls on Gemini to break it into subtopics, then issues several related searches simultaneously on the user’s behalf (Search Engine Land, 2025 (https://searchengineland.com/guide/query-fan-out)). A simple question might generate a handful of sub-queries; a complex or comparative one can generate well over a dozen. **Key takeaway:** You can win a citation by being the best answer to one of the 8 to 12 sub-queries fan-out generates, even if you do not rank for the exact phrase the person typed. The literal query is no longer the only target. ### Does ranking number one on Google guarantee an AI Overview citation? No, and the guarantee has been getting weaker. Across a large-scale study, the share of AI Overview citations that also ranked in the traditional top 10 fell from 76% to 38% within about seven months, meaning a page can now be the cited source without cracking the classic first page at all. #### How much has the overlap between rankings and citations changed? Substantially. Ahrefs’ original analysis of 1.9 million AI Overview citations found 76% also ranked in the top 10, with a median position of 2 (Ahrefs, 2025 (https://ahrefs.com/blog/search-rankings-ai-citations/)). A follow-up study across 863,000 keywords and 4 million AI Overview URLs found that figure had dropped to just 38%, with the remaining citations split almost evenly between pages ranking 11 to 100 (31.2%) and pages ranking beyond position 100 (31.0%) (Ahrefs, 2026 (https://ahrefs.com/blog/ai-overview-citations-top-10/)). Ahrefs notes the shift lines up with Google rolling AI Overviews onto Gemini 3 globally in January 2026. **Key takeaway:** Nearly two in three AI Overview citations now come from outside the traditional top 10. Rank tracking alone will not tell you whether you are showing up in the answer. ### How is an AI Overview citation different from a normal top-10 result? The unit Google is scoring shrinks from the page to the passage, and the target multiplies from one query to a cluster of sub-queries. A page ranking 40th for your target keyword can still win the citation if it is the clearest, most current answer to one of the sub-queries fan-out generated, which is precisely the class of page the 62% figure above accounts for. #### Does content freshness matter more for AI Overviews than classic search? Yes, measurably. Ahrefs analysed roughly 17 million citations across AI Overviews and major AI assistants and found AI-cited content is 25.7% fresher than what shows up in ordinary organic results, with roughly half of cited pages published or meaningfully updated within the prior 13 weeks (Ahrefs, 2025 (https://ahrefs.com/blog/do-ai-assistants-prefer-to-cite-fresh-content/)). A visible last-updated date and an actual recent edit both help more here than they do for classic rankings. #### What if I do not rank in the top 10 at all? You are not automatically out of the running. Because citations now spread across the full ranking spectrum rather than clustering near position one, the more useful question is not "where do I rank for this keyword" but "which of the sub-queries around this topic do I answer better than anyone else." ### What should you actually do to win Google AI Overview citations? Treat the literal search term as one entry point among several, not the whole target, and put the same effort into extractability and freshness that the research rewards. - Answer the surrounding questions, not just the target keyword. List the sub-questions a buyer would ask around your topic and give each one a direct, self-contained answer, the same discipline covered in how AI decides who to cite (/blog/how-ai-decides-who-to-cite). - Keep your highest-value pages genuinely current, not just re-dated. Update the facts, not only the timestamp, since freshness is a scored signal here. - Open every section with the answer in the first sentence or two. Fan-out means Google is matching passages against many sub-queries at once, so a passage that only makes sense in context of the whole page will not survive extraction. - Check that Google-Extended, the crawler behind Google’s AI features, is not blocked in your robots.txt. The check is the same idea covered for ChatGPT’s crawler in why doesn’t ChatGPT recommend my business (/blog/why-doesnt-chatgpt-recommend-my-business), just a different bot name. - Track citations for the query cluster around your target term, not only the exact phrase, since that cluster is what fan-out is actually scoring you against. ### How do you know if it is working? Google does not give you a chat box to prompt-test the way ChatGPT or Claude does, so proving an AI Overview citation takes a different kind of evidence than asking a model directly. Outercite reads the Overview panel Google actually rendered for your query, then runs it through the same judge and verifier models that check every other engine: a candidate citation is confirmed or disputed by an independent second model before it counts. Alongside that, the pipeline actively prompts ChatGPT, Claude, Gemini, Perplexity, Grok and DeepSeek. Two different acquisition methods, one standard of proof. For Google specifically, we also watch the other side of the evidence: real visits. Outercite’s CiteTrace detects traffic attributed to Google AI Overviews, via the Google-Extended crawler and AI Overview referral patterns, and binds those visits back to your existing citation records. Read together with the panel itself, that turns "we think this page might be cited" into "this citation is sending real people to your site." **Key takeaway:** Reused ranking systems plus query fan-out means a top-10 rank is a strong start, not proof. Track the sub-query cluster, keep pages genuinely fresh, and confirm the citation against the Overview Google actually rendered, plus the traffic it sends. ### Sources - AI Overviews and AI Mode share Search’s core ranking and quality systems with no additional requirements to appear: Google Search Central, 2026. Ranking systems guide (https://developers.google.com/search/docs/appearance/ranking-systems-guide). - Query fan-out mechanics and Elizabeth Reid’s Google I/O 2025 remarks on breaking questions into subtopics: Search Engine Land, 2025. Guide (https://searchengineland.com/guide/query-fan-out). - The original 76% top-10 overlap across 1.9 million AI Overview citations, median position 2: Ahrefs, 2025. Study (https://ahrefs.com/blog/search-rankings-ai-citations/). - The drop to 38% top-10 overlap across 863,000 keywords and 4 million AI Overview URLs, with the 31.2%/31.0% position split: Ahrefs, 2026. Study (https://ahrefs.com/blog/ai-overview-citations-top-10/). - AI-cited content 25.7% fresher than organic results, roughly half of citations under 13 weeks old, across 17 million citations: Ahrefs, 2025. Study (https://ahrefs.com/blog/do-ai-assistants-prefer-to-cite-fresh-content/). ## Why doesn't ChatGPT recommend my business? URL: https://outercite.com/blog/why-doesnt-chatgpt-recommend-my-business Author: Arno Verburg, Founder. Published July 20, 2026. You have asked ChatGPT the question your own customer would ask, your business did not come up, and a competitor did. That is a diagnosable problem, not bad luck. In almost every case it comes down to one of four things: the crawler cannot reach you, the crawler cannot read you, your content has nothing worth extracting, or nobody besides you is saying you exist. This is the troubleshooting version. For the general mechanics of how engines choose who to cite, see how AI decides who to cite (/blog/how-ai-decides-who-to-cite), and for the ChatGPT-specific do-this-now checklist, see how to get cited by ChatGPT (/blog/how-to-get-cited-by-chatgpt). Here we work backwards from the symptom to the cause. ### What are the four reasons ChatGPT skips a business? A blocked crawler, a JavaScript-only site, thin or unstructured content, and a lack of third-party presence account for almost every case of a business ChatGPT should plausibly know about but never names. They stack in roughly this order of severity: the first two make you invisible outright, the last two make you visible but unconvincing. - Blocked crawler. Your robots.txt disallows the bot ChatGPT search actually uses. - JavaScript wall. Your content only renders client-side, and the crawler never executes the script that builds it. - Thin or unstructured content. There is nothing extractable: no direct answers, no specifics, nothing to quote. - No third-party presence. The only place that says you exist is your own website. **Key takeaway:** The first two are technical and binary: fixed or not fixed. The last two are about substance and trust, and they compound over months, not days. ### Reason 1: a blocked crawler #### How do I check if I have blocked ChatGPT's crawler? Open yourdomain.com/robots.txt and look for a Disallow rule under user-agent OAI-SearchBot or user-agent *. OAI-SearchBot is the crawler ChatGPT search actually uses to surface live pages; OpenAI states plainly that sites blocking it will not appear in ChatGPT search results (OpenAI, 2026 (https://developers.openai.com/api/docs/bots)). If it is disallowed, this is your answer, and nothing else on this page matters until you fix it. #### Is OAI-SearchBot the same crawler as GPTBot? No, and the distinction matters. GPTBot crawls for model training; OAI-SearchBot crawls for live ChatGPT search results, and OpenAI documents them as separate user-agents with independent rules (OpenAI, 2026 (https://developers.openai.com/api/docs/bots)). You can block GPTBot to opt out of training while leaving OAI-SearchBot allowed, so you stay eligible to be cited. Many site owners block both by copying a generic "block all AI bots" snippet, which also blocks the one bot that would have surfaced them. Blocking is more common than most businesses assume. An Ahrefs analysis of robots.txt files across roughly 140 million websites found GPTBot blocked by 5.89% of sites overall, rising to 7.3% once subdomains are included, with ClaudeBot’s block rate growing by around a third over the prior year as more sites copy each other’s blocklists (Ahrefs, 2025 (https://ahrefs.com/blog/ai-bot-block-rates/)). If your developer, agency or a security plugin ever added a blanket AI-bot block, check it named the right bot. - Check robots.txt for a Disallow rule against OAI-SearchBot specifically, not just GPTBot. - If you use a security plugin, CDN rule or WAF, check it is not silently blocking AI user-agents at the network level, which robots.txt alone will not show you. - Confirm your key pages are not behind a login wall or an interstitial that blocks any automated visitor. ### Reason 2: a JavaScript wall #### Can ChatGPT read a JavaScript-rendered website? Not reliably. A joint analysis of GPTBot’s requests across Vercel’s network found it fetched JavaScript files on roughly 11.5% of visits but found no evidence it ever executed any of them, and the same pattern held for Claude’s and other major AI crawlers (Vercel and Merj, 2024 (https://vercel.com/blog/the-rise-of-the-ai-crawler)). If your page is empty HTML until a script runs, the crawler sees the empty HTML. This is easy to check yourself. Right-click your homepage, choose "View Page Source" (not "Inspect"), and read what is actually there before any script runs. If your business name, services and location are not in that raw HTML, an AI crawler cannot see them either, no matter how the page looks in a browser. - Server-render or statically generate the pages you most want cited, at least the core content: what you do, who for, and where. - If a full rewrite is not realistic, add server-rendered fallback content for the sections that matter most, even a plain-HTML summary block. - Re-check with View Page Source after any fix, since a JavaScript framework can silently reintroduce the problem on the next deploy. ### Reason 3: thin or unstructured content #### What does "not extractable" actually mean? It means a model could not lift a clean, self-contained answer out of your page even if it wanted to. Vague marketing copy, paragraphs that bury the point three sentences in, and pages with no concrete facts, numbers or specifics give the model nothing to quote or cite with confidence. This is measured, not a matter of taste. The Princeton GEO study found that adding quotations, specific statistics and cited sources lifted a page’s visibility in generative answers by roughly 28 to 41% each, the three strongest levers it tested across a 10,000-query benchmark, while keyword stuffing, the old SEO reflex, did almost nothing (Aggarwal et al., arXiv, 2023 (https://arxiv.org/abs/2311.09735)). **Key takeaway:** A page can be fully crawlable and still get skipped if it has nothing specific enough to quote. Extractability is a writing problem as much as a technical one. - Open each section with a direct answer in the first sentence or two, not three paragraphs of throat-clearing first. - Replace vague claims ("fast, reliable service") with specifics (what, how long, how much, since when). - Add a short FAQ where each question is phrased the way a real customer would ask it, with a self-contained answer. - Show a visible last-updated date; freshness is one of the signals engines weight when choosing between similar sources. ### Reason 4: no third-party presence #### Why does it matter what other sites say about me? Because engines trust corroboration over self-description. An analysis of over a million AI prompts found around 85% of citations came from earned, third-party sources rather than a brand’s own website (Muck Rack, 2026 (https://muckrack.com/blog/what-is-ai-reading-may-2026)). If the only place describing your business is your own site, you are asking the model to take your word for it, and it usually will not. This is the slowest fix of the four and the one most businesses skip, which is exactly why it is the highest-leverage one left once the technical issues are sorted. We cover which third-party sources ChatGPT leans on most, and how to earn a place in them, in how to get cited by ChatGPT (/blog/how-to-get-cited-by-chatgpt); do not rebuild that work here, just start it. ### A 10-minute self-diagnosis, in order Work top to bottom. Each step only matters if the ones above it pass, since a blocked crawler makes everything below it irrelevant. 1. Check robots.txt at yourdomain.com/robots.txt for a Disallow rule against OAI-SearchBot or user-agent *. 2. View Page Source (not Inspect) on your most important page and confirm your core content is in the raw HTML, not injected by a script. 3. Read your homepage as a stranger would: does the first paragraph state plainly what you do, for whom, and where, with specifics? 4. Search your business name plus your category. If your own website is the only result, you have no third-party presence yet. 5. Ask ChatGPT the exact question your customer would ask, note who it names instead of you, and check whether those competitors clear steps 1 to 4 more convincingly than you do. ### How do you know the fix worked? Run the same prompt again days or weeks later, not minutes later. The same question can return a different answer run to run, so one retry proves nothing; you need a pattern across repeated checks before you trust a change actually moved the needle. This is what Outercite automates. Our site audit checks whether your robots.txt is inviting or blocking the crawlers that matter, and our citation tracking repeatedly asks the prompts that matter for your business across ChatGPT, Claude, Gemini, Perplexity, Grok and DeepSeek. Every candidate citation is analysed by a judge model, then independently confirmed or disputed by a second verifier model, so a hallucinated mention is never counted as a real one. **Key takeaway:** Fix the technical wall first, then the substance, then the third-party proof. Measure the pattern over repeated checks, not a single screenshot, because that is the only way to know which fix actually worked. ### Sources - OAI-SearchBot as the crawler behind ChatGPT search results, distinct from the GPTBot training crawler: OpenAI crawler documentation. Docs (https://developers.openai.com/api/docs/bots). - GPTBot blocked by 5.89% of roughly 140 million websites (7.3% including subdomains), ClaudeBot block-rate growth: Ahrefs, 2025. Study (https://ahrefs.com/blog/ai-bot-block-rates/). - GPTBot fetching but not executing JavaScript on roughly 11.5% of requests: Vercel and Merj, 2024. Analysis (https://vercel.com/blog/the-rise-of-the-ai-crawler). - The 28 to 41% visibility lift from quotations, statistics and cited sources, and the weak effect of keyword stuffing: Aggarwal et al., GEO: Generative Engine Optimization, KDD 2024. arXiv (https://arxiv.org/abs/2311.09735). - The 85% earned-media citation share across more than one million AI prompts: Muck Rack, 2026. Study (https://muckrack.com/blog/what-is-ai-reading-may-2026). ## How do I get my business cited by ChatGPT? URL: https://outercite.com/blog/how-to-get-cited-by-chatgpt Author: Arno Verburg, Founder. Published July 13, 2026. A customer in your suburb opens ChatGPT and asks for the best plumber, accountant or cafe nearby. It names three businesses and the customer picks one, without ever seeing a list of links. If you were not named, you were not in the running. This guide is the practical, do-this-now version for an Australian small business owner who wants to be one of the three. It is deliberately ChatGPT-specific. If you want the general mechanics of how every engine chooses who to cite, read how AI decides who to cite (/blog/how-ai-decides-who-to-cite) and the broader playbook for earning citations (/blog/get-cited-by-ai). Here we stick to ChatGPT and the moves that matter for it. ### How does ChatGPT decide which businesses to cite? When ChatGPT answers a question that needs current facts, it searches the live web, retrieves a handful of pages, and writes an answer over the ones it trusts most. It is not recalling you from training. It is picking sources in the moment, favouring pages that answer the exact question directly, clearly and with visible credibility. **Key takeaway:** ChatGPT does not remember your business. For anything current, it retrieves and cites live pages. Your job is to be one of the pages it retrieves and trusts, on the web today. #### Does ChatGPT search use Google or Bing? Primarily Bing, plus OpenAI’s own crawl. One analysis of about 100 queries found that 87% or more of ChatGPT search citations matched Bing’s top organic results, versus a 56% match with Google (Seer Interactive, 2025 (https://www.seerinteractive.com/insights/87-percent-of-searchgpt-citations-match-bings-top-results)). The practical takeaway: if you rank well in Bing, you are far likelier to be retrieved, so do not ignore Bing Webmaster Tools the way most Australian businesses do. #### Can ChatGPT cite my site if I have blocked its crawler? No. ChatGPT search uses a dedicated crawler, OAI-SearchBot, to surface sites in its answers. OpenAI states that sites blocking it will not appear in ChatGPT search results (OpenAI, 2026 (https://developers.openai.com/api/docs/bots)). It is separate from GPTBot, which is for model training, so you can allow search visibility while still opting out of training. Check your robots.txt before anything else. ### Why does third-party presence matter more than my own website? Because ChatGPT trusts corroboration over self-description. It leans heavily on a small set of reference, community and review sites rather than the businesses’ own pages. Studies of the domains it cites most put Reddit and Wikipedia at the top by a wide margin, followed by retail and publisher sites. Being talked about in those places beats simply talking about yourself. **Key takeaway:** If your whole ChatGPT plan is to publish more pages on your own site, you are fighting for the smallest slice of the answer. Most of the real work happens off your domain. #### Which websites does ChatGPT cite most? Reddit and Wikipedia dominate, then large retail and publisher sites. An Ahrefs analysis of 9.6 million ChatGPT queries ranked Reddit first, then Wikipedia, Amazon, Forbes and Business Insider by how often they appear in responses (Ahrefs, 2025 (https://ahrefs.com/blog/most-cited-domains-in-chatgpt/)). A larger study across 100 million-plus citations found the same two sources on top, but also that their shares swing sharply month to month (Semrush, 2025 (https://www.semrush.com/blog/most-cited-domains-ai/)). - Community threads. Reddit is the single most-cited source, so genuine, helpful participation in the subreddits for your category and city matters. - Reference pages. Wikipedia turns up constantly; an accurate entry for a notable business or its category is trusted heavily. - Review and directory sites. The trusted places buyers compare businesses in your category, including local Australian directories. - Publisher and roundup pages. Best-of lists and press coverage that name you alongside competitors. #### How does an Australian small business get onto those sources? Start local and earned, not paid. Claim and complete your Google Business Profile, get listed accurately in the directories and review sites for your trade, and gather real reviews with your name, service and suburb in them. Answer questions honestly in the relevant Reddit and community threads. These are the corroborating sources ChatGPT already pulls from for local queries. ### How should I structure my own pages so ChatGPT can extract them? Write self-contained answer blocks that survive being lifted out with no surrounding context. Open each section with a direct answer of about 40 to 60 words, phrase headings as the questions people actually ask, and keep one idea per passage. ChatGPT cites the passage, not the page, so the passage has to stand on its own. Credibility is measurable, not a matter of taste. The Princeton GEO study found the right moves can lift a source’s visibility in generative answers by up to 40%, with the biggest gains from adding quotations, statistics and cited sources; keyword stuffing was one of the weakest tactics tested (Aggarwal et al., arXiv, 2023 (https://arxiv.org/abs/2311.09735)). Show your working with real numbers and references, and do not pad with repeated keywords. #### What content format does ChatGPT prefer? Direct, structured and current. Lead with the answer, use question-shaped headings, add an FAQ where each answer stands alone, and use lists for options and steps. Show a visible last-updated date and name a real author with credentials. The same discipline underpins answer engine optimisation (/blog/what-is-aeo) generally, and ChatGPT rewards it as much as any engine. ### What is the do-this-now checklist? You can make real progress in an afternoon. Work top to bottom: fix the technical blockers first, because if OAI-SearchBot cannot read you, nothing else you do can help. Then earn off-site presence and sharpen your own pages. - Check your robots.txt does not block OAI-SearchBot, and confirm your key pages are not behind a login or heavy JavaScript. - Claim and fully complete your Google Business Profile with correct services, hours and suburb. - Get listed accurately in the review and directory sites for your trade, and ask happy customers for reviews that mention your service and location. - Verify your site in Bing Webmaster Tools and fix anything stopping Bing from indexing you, since ChatGPT leans on the Bing index. - Rewrite your most important pages to open each section with a direct answer, with question-shaped headings and a short FAQ. - Add a visible last-updated date and a named author with real credentials. - Ask ChatGPT the question your best customer would ask, and note who it names instead of you. That is your starting line. ### How do I tell if it is working? Measure repeatedly, across the prompts that matter, not once. The same question returns different answers run to run, so a single screenshot proves nothing. Track how often ChatGPT names you, for which questions, and against which competitors, over weeks. A pattern over time is proof; a one-off answer is noise. This is what Outercite does. We track how AI search engines cite businesses across ChatGPT, Claude, Gemini, Perplexity, Grok and DeepSeek. Every candidate citation is analysed by a judge model, then confirmed or disputed by an independent verifier model before it counts, so a hallucinated mention is never scored as a real citation. **Key takeaway:** Getting cited by ChatGPT is not luck. Unblock the crawler, earn presence in the sources it trusts, make your pages extractable, then measure the pattern over time. ### Sources - ChatGPT search citation overlap with Bing (87%-plus match versus 56% for Google, roughly 100 queries): Seer Interactive, 2025. Analysis (https://www.seerinteractive.com/insights/87-percent-of-searchgpt-citations-match-bings-top-results). - OAI-SearchBot surfaces sites in ChatGPT search and is separate from the GPTBot training crawler: OpenAI crawler documentation. Docs (https://developers.openai.com/api/docs/bots). - Most-cited ChatGPT domains ranked (Reddit, Wikipedia, Amazon, Forbes, Business Insider) from 9.6 million queries: Ahrefs, 2025. Study (https://ahrefs.com/blog/most-cited-domains-in-chatgpt/). - Reddit and Wikipedia leading but volatile across 100 million-plus citations: Semrush, 2025. Study (https://www.semrush.com/blog/most-cited-domains-ai/). - The up-to-40% visibility lift and the strongest tactics (quotations, statistics, cited sources) versus weak keyword stuffing: Aggarwal et al., GEO: Generative Engine Optimization, KDD 2024. arXiv (https://arxiv.org/abs/2311.09735). ## What is generative engine optimisation (GEO)? URL: https://outercite.com/blog/what-is-generative-engine-optimisation Author: Arno Verburg, Founder. Published July 8, 2026. Generative engine optimisation, usually shortened to GEO, is the newest label for an old goal: being the source an answer is built on. The difference is the surface. Instead of a ranked list of links, the answer is a written paragraph that a generative engine has synthesised, and GEO is the work of getting your content used inside it. **Key takeaway:** GEO is the practice of structuring and placing your content so generative engines cite and recommend it inside their generated answers, rather than merely ranking it in a list of links. ### What is generative engine optimisation? GEO is the discipline of improving how often, and how favourably, a generative engine cites your content when it composes an answer. A generative engine is any AI system that retrieves sources and writes a synthesised response over them, such as ChatGPT, Claude, Gemini, Perplexity, Grok, Google AI Overviews or DeepSeek. GEO optimises for that citation, not for a rank. #### What counts as a generative engine? A generative engine is a system that answers a question by retrieving candidate sources and generating prose over them, then citing some of what it used. That covers AI assistants like ChatGPT and Claude, answer engines like Perplexity, and Google AI Overviews sitting on top of classic search. The common thread is that the user reads a written answer, not a page of links. #### Is GEO an SEO replacement? No. GEO sits alongside SEO rather than replacing it. The engines still retrieve live sources, and strong search foundations make you easier to retrieve, so good SEO helps GEO. What changes is the outcome you optimise for and measure: a citation inside an answer, rather than a position in a list. ### GEO vs AEO vs SEO: what is the difference? These three overlap heavily and people use them loosely. The short version: SEO competes for a rank on a results page, while GEO and AEO both compete to be cited inside a generated answer. GEO and AEO are near-synonyms in practice, with GEO the term the research community adopted and AEO the one the marketing world tends to use. - SEO (search engine optimisation) optimises for position in a ranked list of links, and is measured in rankings, clicks and impressions. - AEO (answer engine optimisation) optimises for being cited and recommended inside an AI-generated answer, and is measured in citations and share of voice across engines. - GEO (generative engine optimisation) targets the same outcome as AEO, being used as a source inside a generated response, and is the label coined by the original academic research. **Key takeaway:** Treat GEO and AEO as the same job under two names. If you have read our explainer on what AEO is (/blog/what-is-aeo), you already understand GEO. #### Is GEO the same as AEO? For almost all practical purposes, yes. Both describe getting your content cited inside AI-generated answers rather than ranked in a list. GEO comes from academic research and AEO from the marketing community, but the moves, the metrics and the goal are the same. We cover the answer-engine framing in what is AEO (/blog/what-is-aeo). #### How is GEO different from SEO? SEO wins a position in a list of ten links, where the click is the prize. GEO wins a citation inside a single written answer, where being the source is the prize. They share foundations like clear structure and genuine authority, but they measure different outcomes. We break the split down in AEO vs SEO (/blog/aeo-vs-seo). ### Where did the term GEO come from? The term was introduced in a 2023 research paper, "GEO: Generative Engine Optimisation" by Pranjal Aggarwal and colleagues, later accepted to the KDD 2024 conference (arXiv, 2023 (https://arxiv.org/abs/2311.09735)). The authors, from teams including Princeton and Georgia Tech, coined "generative engine" for AI systems that generate answers over retrieved sources, and "GEO" for optimising to be cited by them. #### Who coined the term generative engine optimisation? The paper by Aggarwal, Murahari, Rajpurohit, Kalyan, Narasimhan and Deshpande introduced both the concept of a generative engine and the GEO acronym (arXiv, 2023 (https://arxiv.org/abs/2311.09735)). It predates most of the marketing writing on the topic, which is why GEO is the term with a citable academic origin, while AEO grew up in industry blogs. ### What did the Princeton GEO study find? It found that GEO is measurable and that a handful of tactics move the needle. Testing across a benchmark of 10,000 queries, the authors reported that GEO methods can lift a source’s visibility in generative engine responses by up to 40%, with the largest gains from adding quotations, statistics and cited sources (Aggarwal et al., arXiv, 2023 (https://arxiv.org/abs/2311.09735)). #### Which GEO tactics worked best? On the study’s position-adjusted word-count metric, the three strongest tactics were adding quotations, adding statistics and citing sources, each lifting visibility by roughly 28 to 41% over the no-optimisation baseline (Aggarwal et al., arXiv, 2023 (https://arxiv.org/abs/2311.09735)). Quotations were the single biggest lever. The through-line is credibility: engines favour content that shows its working with quotes, numbers and references. #### Does keyword stuffing help with GEO? No. The study found that keyword stuffing, the old SEO reflex of packing in relevant terms, offered little to no improvement in generative engine responses and was among the weakest tactics tested (Aggarwal et al., arXiv, 2023 (https://arxiv.org/abs/2311.09735)). Repetition does not persuade a model that synthesises meaning, so the effort is better spent on substance. ### How do you start with GEO? Start by measuring where you already stand, then work on structure, credibility and off-site presence. The research points at concrete moves: write self-contained answers, back claims with quotes and statistics, cite your sources, and earn accurate mentions in the third-party places engines already trust for your category. - Structure for extraction. Open each section with a direct answer, use question-shaped headings, and keep one idea per passage so it survives being lifted out. - Show credibility. Add real quotations, specific statistics and cited sources, the three tactics the study found most effective. - Skip the old reflexes. Do not keyword-stuff; the research shows it adds little and can read as spam. - Build off-site presence. Earn an accurate, current presence in the review sites, roundups and communities the engines retrieve from. - Measure across engines, repeatedly. The same prompt returns different answers run to run, so a single check is noise. Track citations over time. ### How Outercite measures GEO Outercite tracks how AI search engines cite businesses across ChatGPT, Claude, Gemini, Perplexity, Grok and DeepSeek. Rather than trusting a single answer, every candidate citation is checked by a judge model and then confirmed or disputed by an independent verifier model before it counts, so a hallucinated mention is not scored as a real citation. That two-model check is the point. GEO only compounds if you can tell a genuine citation from noise and prove a change moved the number, which is the discipline the research rewards and the guesswork it replaces. ### Sources - Origin of the term, the 10,000-query benchmark, the up-to-40% visibility lift, the per-tactic ranking (quotations, then statistics, then cited sources, each lifting visibility by roughly 28 to 41% on the position-adjusted word-count metric) and the keyword-stuffing finding: Aggarwal et al., "GEO: Generative Engine Optimisation", KDD 2024. arXiv (https://arxiv.org/abs/2311.09735). ## How AI decides who to cite (and how to become the answer) URL: https://outercite.com/blog/how-ai-decides-who-to-cite Author: Arno Verburg, Founder. Published June 29, 2026. Your best customer just asked ChatGPT for a recommendation in your category. It named three businesses. You were not one of them. They did not pay to be there. They did not rank first on Google. The model simply chose them as the answer, and the customer never saw a list of ten blue links to scroll. This is the new front door, and most businesses are still polishing the old one. I run Outercite, where we track how AI search engines cite Australian businesses across ChatGPT, Claude, Gemini, Perplexity, Grok and DeepSeek. I have spent a lot of hours watching who gets named and who gets skipped, and the pattern is not random. It is learnable. This is the field guide I wish existed when I started. ### The one shift that reorganises everything Stop trying to rank a page. Start trying to be an extractable answer. Traditional SEO gets you ranked on a results page. AI search gets you cited inside an answer. They share a foundation, but they reward different things: SEO optimises for the click, AI search optimises for the reuse. The unit is no longer the page. It is the passage. A page sitting on result two can be cited over the page on result one if its answer is cleaner and easier to lift. Three words people use interchangeably are not the same thing: - Ranked. You sit high on a Google results page. A separate game. - Mentioned. Your name appears somewhere in an AI answer. Nice, and passive. - Cited. You are the source the answer was actually built on. **Key takeaway:** Citation is the one that moves customers, and the one most AI visibility tools quietly fail to measure. Counting mentions is easy. Counting real citations is hard. ### The secret that should reorder your strategy You get cited far more often from other people’s websites than from your own. In one analysis of more than a million AI prompts, around 85% of citations came from earned, third-party sources, not the brands’ own websites. That is roughly five to six times more likely to be cited off your own site than on it, and we see the same pattern across the Australian businesses we track. So if your whole AI search plan is to write more content on your own site, you are fighting for the smallest slice of the answer. The engines trust corroboration over self-description, so most of the real work happens off your domain, in what I call the citation supply chain: - Reference and community sources. Wikipedia and Reddit turn up constantly. - Review and comparison sites. The trusted directory for your category. - Best-of and roundup pages. The third-party lists buyers compare against. - YouTube. Frequently cited by Google’s AI Overviews for how-to and product questions. Find where the engines that matter to you already pull their answers for your category, then earn an accurate, current presence in those exact places. Most businesses have never looked. That is the gap. ### How engines actually choose who to cite They read a page top to bottom, lean heavily on headings to understand intent, and prefer sources that answer directly, clearly, completely and credibly. Underneath a cited answer is a short pipeline: the engine reads the intent, retrieves candidate sources by searching live, selects the few it trusts most, and cites the ones it leaned on. The answers that carry citations are almost always retrieval answers, and retrieval is the part you can actually influence. ### What actually works: three levers #### 1. Structure: be extractable Make every section a self-contained answer block that works if it is lifted out with no surrounding context. - Open each section with the answer, in about 40 to 60 words, then expand. - One idea per section, one idea per paragraph. Keep paragraphs to two to four sentences. - Write your headings the way a person phrases the question. - Use lists for options and tables for comparisons. - Add an FAQ with self-contained answers. #### 2. Authority: be citable This is measured, not a matter of taste. The Princeton GEO study found the right moves can lift a source’s visibility in AI answers by up to 40%, with the biggest gains from citing your sources, adding specific statistics, and adding expert quotations. Keyword stuffing, the old SEO reflex, was one of the weakest tactics it tested. Freshness is weighted heavily, so show a visible last-updated date, and name your authors with real credentials. #### 3. Presence: be where AI looks This is the third-party supply chain from earlier, and it is the lever with the most headroom precisely because it is the most work. Map the sources, earn the placements, keep them current. If you do only one new thing this quarter, do this. ### The plumbing nobody checks Four technical moves decide whether the engines can read you at all. They take an afternoon and almost nobody does them. - AI bot access. If GPTBot, PerplexityBot, ClaudeBot or Google-Extended are blocked in your robots.txt, that engine literally cannot cite you. Check this first. - Schema markup. Structured data tells engines what your content is without guessing. - An llms.txt file. A clean map of what you do for AI systems. - A pricing.md file. AI agents are becoming the ones comparing products for a buyer. If your pricing is behind JavaScript or a contact-sales wall, they skip you. ### How do you know if it is working You measure at the passage level, repeatedly, across engines, because a single check is noise. The trap almost everyone falls into is to run one prompt, screenshot the answer, and treat it as proof. But the same prompt returns different answers run to run and engine to engine. **Key takeaway:** Proof is a pattern measured over time, not a screenshot. And you have to separate a real citation, where the engine used you as a source, from a hallucinated mention. At Outercite we verify every citation with two independent models before it counts. The metrics that actually reflect AI visibility are citation frequency, share of voice against your competitors, and query coverage across the related questions buyers ask. ### Turn it into a system: See, Understand, Act, Prove AI search visibility is not a project you finish. It is a loop you run. - See. Where you stand today: coverage, citation rate, share of voice. - Understand. Why a competitor is cited and you are not: entities, authority, third-party trust, freshness. - Act. The highest-leverage moves first, sequenced by impact. - Prove. Track each change, tie it to citation movement, and repeat. > Run it every quarter and it compounds. That is the difference between a flurry of activity that moves nothing, and a system that keeps making you the answer. ### If you are an agency, this is a service There are no established AI search agencies yet, and no one has a ten-year head start. The agency that walks into a client meeting with a clear method and a plain-English brief wins the retainer. Package the loop into a productised service: an audit that shows the client where they stand, a one-page brief that turns the gaps into a scoped, priced ask, and a monthly report that proves it worked. ### The honest part This takes about 90 days to compound, especially the off-site work. Leading signals like citation rate move first, with traffic and revenue following. Anyone promising instant AI visibility is selling you something. The good news is that the field is young, the moves are knowable, and most of your competitors are not doing them yet. You do not need any tool to start. Open ChatGPT, ask it the question your best customer would ask, and see whether it names you. That answer is your starting line. ### Sources - The 40% visibility lift and the top tactics: Aggarwal et al., GEO: Generative Engine Optimization, KDD 2024. arXiv (https://arxiv.org/abs/2311.09735). - The earned-media citation share: Muck Rack, analysis of over one million AI prompts. Study (https://muckrack.com/blog/what-is-ai-reading-may-2026). - AI crawler controls: official docs for OpenAI GPTBot (https://platform.openai.com/docs/bots), PerplexityBot (https://docs.perplexity.ai/guides/bots), Anthropic ClaudeBot (https://support.anthropic.com) and Google-Extended (https://developers.google.com/search/docs/crawling-indexing/overview-google-crawlers). ## What is Answer Engine Optimisation (AEO)? A 2026 guide URL: https://outercite.com/blog/what-is-aeo Author: Celia Young, Head of Marketing. Published June 18, 2026. When someone asks ChatGPT "what is the best AI visibility platform," the answer they get is a recommendation, not a list of ten blue links. Answer Engine Optimisation (AEO) is the practice of making sure your brand is the one named in that answer. **Key takeaway:** AEO is the discipline of improving how often, and how favourably, AI engines cite and recommend your brand inside generated answers. It is also called generative engine optimisation (GEO). ### Why AEO matters now Buyers are changing how they discover and decide. Instead of scanning a results page, a growing share now ask an AI assistant directly and act on the single answer it gives. That answer is generated, synthesised from many sources, and it usually names a small handful of brands. If you are not one of them, you are invisible at the exact moment of decision. This is a structural shift, not a fad. The engines that matter, ChatGPT, Claude, Gemini, Perplexity, Grok and DeepSeek, are becoming the front door to research, comparison and purchase. AEO is how you make sure that front door opens to you. ### What AEO actually optimises for Traditional SEO optimises for position in a ranked list. AEO optimises for three different things: - Citation. Is your brand mentioned at all when buyers ask the questions that matter in your category? - Recommendation. When you are mentioned, are you recommended favourably, or described as an also-ran? - Coverage. Across how many engines and how many high-intent prompts do you appear, versus your competitors? ### How to approach AEO Good AEO is a loop, not a one-off audit. You measure where you stand across engines, find the prompts where rivals are cited and you are not, decide which gaps are worth closing, do the work, and then measure whether your citations actually moved. > Monitoring tells you where you rank today. AEO is about knowing what to do next, and what it is worth. That last step is where most teams fall down. It is easy to generate a chart of where you stand. It is much harder to know which move will move you, and to prove it did. This is exactly the gap Outercite is built to close: forecasting the uplift of each move before you invest, and measuring the lift after it ships. ### Getting started Start by tracking the buying and comparison prompts that decide your category across the engines your buyers actually use. From there, prioritise the citation gaps your competitors are winning, and treat each fix as a measurable move rather than a guess. AEO rewards teams who run it as a disciplined, evidence-led practice. ## AEO vs SEO: what changes when buyers ask AI URL: https://outercite.com/blog/aeo-vs-seo Author: Marcus Lai, Head of Search. Published June 11, 2026. If you have spent a decade getting good at SEO, AEO can feel both familiar and unsettling. The good news: your instincts about quality content and authority still matter. The catch: what you measure, and how you win, are genuinely different. ### The core difference SEO improves your position in a list of links on a results page. AEO improves whether an AI engine cites and recommends you inside a written answer, where there is often no list of links at all. **Key takeaway:** SEO competes for a rank. AEO competes to be the recommendation. One is a position in a list; the other is a sentence in an answer. ### Where they overlap Strong, well-structured, genuinely useful content helps both. Clear authority signals, accurate information and good site structure make you easier to rank and easier to cite. If your SEO foundations are solid, you are not starting AEO from zero. ### Where they diverge - Measurement. SEO tracks rankings, clicks and impressions. AEO tracks citations, share of voice across engines, and how favourably you are described. - The surface. SEO targets a single results page per query. AEO spans many engines, each of which synthesises answers differently. - The unit of work. SEO often optimises a page for a keyword. AEO optimises your presence across the prompts and comparisons buyers actually ask. - Volatility. AI answers can shift as models update, so AEO rewards teams who detect prompt shifts early rather than reacting late. ### How to run both Do not abandon SEO. Treat AEO as a parallel practice that reuses your content strength but measures a new outcome. Track AI citations alongside your rankings, watch which competitors win the answers you want, and forecast which moves will actually shift your citations before you spend. > The teams that win the next decade will not choose between SEO and AEO. They will run both, and measure each on its own terms. The reporting discipline matters most. In SEO you learned to tie work to ranking changes. In AEO, tie each move to a forecasted and then measured change in citations, so brand and content spend is backed by evidence rather than opinion. ## How to get your brand cited by ChatGPT and Perplexity URL: https://outercite.com/blog/get-cited-by-ai Author: Priya Nair, AEO Strategist. Published June 4, 2026. Getting cited by an AI engine is not luck. It follows a repeatable pattern: understand the questions buyers ask, understand why a competitor is named instead of you, and close that gap deliberately. Here is the playbook. ### 1. Map the prompts that decide your category Start with the questions a real buyer asks an AI on the way to a decision: "best tool for X," "X vs Y," "is X good for small teams." These buying and comparison prompts are where citations convert into pipeline. List them before you do anything else. ### 2. See where you stand, across every engine Ask those prompts across ChatGPT, Claude, Gemini, Perplexity, Grok and DeepSeek and record who gets cited. You will usually find you appear strongly on some engines and prompts and are invisible on others. That map is your starting line. **Key takeaway:** You cannot improve what you cannot see. The first real win in AEO is an honest, engine-by-engine picture of where you are cited and where you are not. ### 3. Diagnose why a competitor wins When a rival is named and you are not, there is almost always a reason: a stronger comparison page, more third-party reviews, clearer positioning, or authoritative coverage the model trusts. Treat each gap as a diagnosis, not a mystery. - Is there a comparison or "best of" page the engine is leaning on? - Do you have enough credible third-party validation, like reviews and mentions? - Is your own content clear and structured enough for a model to quote confidently? - Are you positioned for the exact intent of the prompt, or something adjacent? ### 4. Close the gap as a measurable move Pick the gaps worth closing and treat each as a move with an expected return, not a vague content task. Publish the comparison, earn the reviews, sharpen the positioning. The discipline is to know which move is likely to shift your citations before you invest the effort. > Forecast the move, ship the work, then measure whether your citations actually moved. Repeat. That loop is how brands compound their presence in AI answers. ### 5. Watch for prompt shifts AI answers change as models update and as new content enters the graph. The brands that stay cited are the ones that detect those shifts early and respond, rather than discovering months later that a competitor quietly took their spot. Make monitoring continuous, not quarterly. Run this loop consistently and citations stop being something that happens to you and become something you earn on purpose. ## Knowledge base index - Knowledge Base: https://outercite.com/learn (Guides, playbooks and reference docs for brands and agencies tracking how AI engines cite them across seven monitored surfaces.) - Coverage and credits: https://outercite.com/learn/account/coverage-and-credits (Coverage is a wallet of answers, in two pools, allocated each period. Here is what spends it, what happens when it runs out, and how to make it go further.) - Invoices and payment: https://outercite.com/learn/account/invoices-and-payment (Where invoices live, how to change the payment method or billing email, and what happens if a payment fails.) - Manage your subscription: https://outercite.com/learn/account/manage-subscription (Upgrade, downgrade, switch to annual or cancel, and understand what changes immediately versus at the next period.) - Plans and limits: https://outercite.com/learn/account/plans-and-limits (The lanes on sale, what varies between them, and where to check what your own account actually has.) - Seats and team access: https://outercite.com/learn/account/seats-and-team-access (How team members are counted, what each role can do, and how workspace membership decides what somebody sees.) - Agency billing model: https://outercite.com/learn/agencies/core-concepts/agency-billing-model (One plan, one invoice, one pooled allowance. What appears on the bill, and how to track what each client actually costs you.) - Client permissions model: https://outercite.com/learn/agencies/core-concepts/client-permissions-model (Two things control access: a role, which says what somebody can change, and workspace membership, which says whose data they can see.) - Parent orgs vs workspaces: https://outercite.com/learn/agencies/core-concepts/parent-orgs-vs-workspaces (One agency organisation, one workspace per client. What lives at each level, what is shared, and what never crosses between clients.) - White-label reports: https://outercite.com/learn/agencies/core-concepts/white-label-reports (What you can brand as your own, where the branding applies, and the one thing that stays visible by design.) - Handle a client handover: https://outercite.com/learn/agencies/how-to/handle-client-handover (Moving a client between account managers without losing context, access, or the thread of what was already shipped.) - Import prompts in bulk: https://outercite.com/learn/agencies/how-to/import-prompts-in-bulk (There is no CSV upload. The three ways to add many prompts quickly are discovery multi-select, the API, and the MCP connector.) - Onboard a new client: https://outercite.com/learn/agencies/how-to/onboard-a-new-client (A repeatable sequence for getting a client workspace from empty to reporting, with the checks that stop you doing it twice.) - Schedule client reports: https://outercite.com/learn/agencies/how-to/schedule-client-reports (One schedule per client workspace: recipients, cadence, report type, subject, your own commentary, and a shareable link.) - Send the weekly client update: https://outercite.com/learn/agencies/how-to/send-the-weekly-digest (A short weekly email keeps a retainer visible between monthly reviews. Here is what to put in it and what to leave out.) - Agency client onboarding playbook: https://outercite.com/learn/agencies/playbooks/agency-onboarding-playbook (A repeatable five-day sequence from signed to first finding, so every client gets the same start and you stop reinventing it.) - Monthly client review template: https://outercite.com/learn/agencies/playbooks/monthly-client-review-template (A five-section structure for a monthly review that survives scrutiny, built on what the ledger can actually prove.) - Pitching AI visibility to clients: https://outercite.com/learn/agencies/playbooks/pitching-ai-search-to-clients (How to present AI search visibility to a client who has never thought about it, using their own data rather than industry statistics.) - Upselling with forecasts: https://outercite.com/learn/agencies/playbooks/upselling-with-forecasts (How to use a projected range to make an honest case for more scope, without turning a range into a promise.) - Agency quickstart: https://outercite.com/learn/agencies/start-here/agency-quickstart (Set up your parent organisation, add your first client workspace, scope the team correctly, and get reporting running.) - Creating your first workspace: https://outercite.com/learn/agencies/start-here/creating-your-first-workspace (A workspace is one client: their profile, prompts, competitors, citations and changes, isolated from every other client in your account.) - Partner pricing explained: https://outercite.com/learn/agencies/start-here/partner-pricing-explained (An agency is the Scale plan at the partner price: one allowance, one invoice, and clients as workspaces up to the fair-use cap.) - How answers are grounded: https://outercite.com/learn/assistant/how-answers-are-grounded (Every answer is built from counted rows in your own workspace, in phases you can see, with the evidence for each figure one click away.) - Panes and the ladder: https://outercite.com/learn/assistant/panes-and-the-ladder (Every lens in the platform descends the same three steps: the landscape you scan, the subject you focus, and the proof behind that subject.) - Shortcuts and navigation: https://outercite.com/learn/assistant/shortcuts-and-navigation (The Shortcuts palette, the two top-bar menus, and the four keyboard chords that make the assistant fast.) - What you can ask: https://outercite.com/learn/assistant/what-you-can-ask (The classes of question the assistant answers, the panel each one opens, and how to phrase a question so it lands where you want.) - When a number is missing: https://outercite.com/learn/assistant/when-a-number-is-missing (Why the platform says not measured, excluded or unmeasurable instead of showing you a zero, and what each of those states means.) - Contact support: https://outercite.com/learn/help/contact-support (How to reach us, what to include so the first reply is useful, and the things that are faster to fix yourself.) - Frequently asked questions: https://outercite.com/learn/help/faq (Short answers to the questions people ask in their first month: what is measured, how, on which surfaces, and why some numbers are deliberately missing.) - Why does Outercite show fewer citations than a manual search?: https://outercite.com/learn/help/why-fewer-citations-than-google (A text search counts every appearance of your name. Verification counts recommendations. The gap is the whole point of the number.) - Why is my citation rate low?: https://outercite.com/learn/help/why-is-my-citation-rate-low (Work down the causes in the order they actually fail: access, extractability, authority, prompt quality, then content.) - Why is there no data yet?: https://outercite.com/learn/help/why-no-data-yet (A new workspace is empty for a handful of reasons, and all of them are quick to check.) - API access: https://outercite.com/learn/integrations/api-access (Create a key, authenticate with a bearer token, and pull citations, prompts, competitors, changes, forecasts and Trace data into your own tools.) - Google Search Console: https://outercite.com/learn/integrations/google-search-console (What the Search Console connector brings in, what it does not, and what to do with the disagreements between search and AI visibility.) - MCP connector: https://outercite.com/learn/integrations/mcp (Connect your workspace to an MCP-capable AI client so it can read your citations, competitors, changes and forecasts, and act on them.) - Integrations overview: https://outercite.com/learn/integrations/overview (What you can connect: your store, Search Console, Slack, webhooks, the public API, and the MCP connector for AI clients.) - Slack notifications: https://outercite.com/learn/integrations/slack (Route alerts into a Slack channel, and keep the volume low enough that people still read them.) - Webhooks: https://outercite.com/learn/integrations/webhooks (Send events to your own systems as they happen: the event list, the payload, the signature header, and the retry behaviour.) - WebMCP: https://outercite.com/learn/integrations/webmcp (The browser standard that lets a web page hand typed, callable tools to an in-browser AI agent. What it is, where it actually stands, and whether you should implement it yet.) - How verification works: https://outercite.com/learn/intro/how-verification-works (What happens between an AI engine's answer and a counted citation: the prefilter, the judge, the verifier that audits it, and the rules that decide a tie.) - Meet the assistant: https://outercite.com/learn/intro/meet-the-assistant (The assistant is the front door of the platform. Ask a question in your own words and the answer arrives with a panel of evidence beside it.) - What is Outercite?: https://outercite.com/learn/intro/what-is-outercite (A plain-English introduction to AI citations, the seven surfaces Outercite monitors, and what the platform does with what it finds.) - Why AI search matters: https://outercite.com/learn/intro/why-ai-search-matters (Buyers ask AI before they open a search engine, and an AI answer names one or two businesses instead of listing ten. Here is what that changes.) - Changes: https://outercite.com/learn/lenses/changes (The ledger of what you shipped and whether it worked. Every change climbs a ladder of evidence: proposed, attested, observed live, confirmed moved.) - Competitors: https://outercite.com/learn/lenses/competitors (Who gets named instead of you. The landscape ranks the rivals appearing in your answers, and each rival opens a head to head built from the answers themselves.) - Industry Pulse: https://outercite.com/learn/lenses/industry-pulse (How every workspace we benchmark is doing, and where you sit against it. Read the caveats before you quote it.) - Intelligence: https://outercite.com/learn/lenses/intelligence (One surface with six lenses: Intelligence, Ground, Where you lose, Engines, Gaps and Explore, plus the forecast record. Scale and Enterprise.) - Platforms: https://outercite.com/learn/lenses/platforms (Your citation rate on each of the seven monitored surfaces, with exclusions reported as exclusions rather than as zeroes.) - Prompts: https://outercite.com/learn/lenses/prompts (The questions you are measured on. Prompts are grouped into topics, checked on every surface your plan covers, and read as a grid rather than a list.) - Proof: https://outercite.com/learn/lenses/proof (The verbatim sentences AI engines wrote about you, with the engine and the date attached. The strongest evidence in the product.) - Readiness: https://outercite.com/learn/lenses/readiness (The grade on your own site: whether engines can reach it, parse it, trust it, and whether they actually cite it. Four levels and four named tiers.) - Sources: https://outercite.com/learn/lenses/sources (The pages and domains AI engines lean on when they answer questions in your category, as a counted graph you can descend to the citing answers.) - The last run: https://outercite.com/learn/lenses/the-last-run (What the agents did overnight: the steps they actually took, what they found, and what they wrote, with each part labelled by how solid it is.) - This week: https://outercite.com/learn/lenses/this-week (The ranked plan: which open change to do first, what it is projected to be worth, and how much of that projection is calibrated against shipped outcomes.) - Trace: https://outercite.com/learn/lenses/trace (Who arrived from an AI answer, which assistant sent them, and what they were reading. The only lens that connects a citation to a real visit.) - Competitors and share of voice: https://outercite.com/learn/organisations/core-concepts/competitors-and-share-of-voice (Share of voice is your portion of the naming across your tracked prompts. Here is how it moves, and what each kind of movement means.) - Forecasts and ranges: https://outercite.com/learn/organisations/core-concepts/forecasts-and-ranges (A forecast is a range, not a number, and its value is the record of what happened last time one was made. How to read it and what not to promise.) - How a citation is counted: https://outercite.com/learn/organisations/core-concepts/how-a-citation-is-counted (What has to be true before a check becomes a counted citation, what the confidence score measures, and how to use it in a decision.) - Intent classification: https://outercite.com/learn/organisations/core-concepts/intent-classification (Every prompt is sorted into one of five intent types before anything is asked, so a buying-intent citation is never averaged with a passing mention in an explainer.) - The six-level stack: https://outercite.com/learn/organisations/core-concepts/the-six-level-stack (The six levels that decide whether AI names you: four on your site, two in the loop you run. They fail bottom up, so fix them in order.) - What is a citation?: https://outercite.com/learn/organisations/core-concepts/what-is-a-citation (A citation is your business being named inside the answer an AI engine writes, whether by name, by URL, by phone number or by product.) - Add a competitor: https://outercite.com/learn/organisations/how-to/add-a-competitor (Add the businesses you sell against, and let the ones the engines actually recommend find you.) - Add prompts: https://outercite.com/learn/organisations/how-to/add-prompts (Three ways to add the questions you are measured on: write them yourself, accept what discovery proposes, or take them from what you are losing.) - Connect Google Search Console: https://outercite.com/learn/organisations/how-to/connect-google-search-console (Link Search Console to see traditional search performance beside your AI citation data, and to spot where the two disagree.) - Export a report: https://outercite.com/learn/organisations/how-to/export-a-report (The four reports, what each is for, how to export one, and how to schedule recurring delivery to the people who need it.) - Install Trace: https://outercite.com/learn/organisations/how-to/install-trace (Put the Trace tag on your site so arrivals from AI answers can be counted, and add the edge worker if you want AI crawler fetches too.) - Invite your team: https://outercite.com/learn/organisations/how-to/invite-your-team (Add teammates, choose a role, and scope an invitation to a single workspace when someone should only see one brand or one client.) - Set up alerts: https://outercite.com/learn/organisations/how-to/set-up-alerts (Choose what you are told about, how fast, and on which channel, then set quiet hours so the important ones still get through.) - Ship and prove a change: https://outercite.com/learn/organisations/how-to/ship-and-prove-a-change (The loop that separates a change you can prove from a change you can only claim: approve it in the ledger, ship it, let it be observed, then let it be confirmed.) - Build AI visibility for a new product: https://outercite.com/learn/organisations/playbooks/launching-a-new-product (A launch playbook that starts before launch: baseline the prompts you intend to win, publish the things engines can quote, then measure the climb.) - Run a quarterly brand health review: https://outercite.com/learn/organisations/playbooks/quarterly-brand-health-review (A repeatable quarterly review built on what can actually be proven: movement, share, the changes you shipped, and what you will do next.) - Recover lost citations: https://outercite.com/learn/organisations/playbooks/recovering-lost-citations (A diagnostic sequence for when engines stop citing you: establish what actually changed, then match the fix to the cause.) - Win local and buying-intent prompts: https://outercite.com/learn/organisations/playbooks/winning-local-buying-intent (A playbook for the two prompt types that sit closest to a purchase: what to track, what to fix, and how to tell whether it worked.) - Organisation quickstart: https://outercite.com/learn/organisations/start-here/org-quickstart (From a new account to a first real answer, then a five-day plan to turn it into a loop you can keep running.) - Reading your first answer: https://outercite.com/learn/organisations/start-here/reading-your-first-answer (What every part of an assistant answer means: the figure in the sentence, the rows in the panel, the basis badges, and the footer that says what was counted.) - Your first prompts: https://outercite.com/learn/organisations/start-here/your-first-prompts (How to choose the questions you want to be the answer to, how many to start with, and why your brand name never belongs in one.) - API v1 reference: https://outercite.com/learn/reference/api-v1 (The public endpoints, their methods, authentication, pagination, errors and rate limits.) - Changelog: https://outercite.com/learn/reference/changelog (Where product changes are published, and how to tell whether a change to the platform affects a number you already reported.) - Data definitions: https://outercite.com/learn/reference/data-definitions (What each metric counts, what it excludes, and over which window. The page to check before quoting a number.) - Glossary: https://outercite.com/learn/reference/glossary (Plain-English definitions for every term the product puts in front of you, and the ones it deliberately avoids.) - System status: https://outercite.com/learn/reference/system-status (How to tell whether something is broken on our side, what a partial degradation looks like, and how to report one.) - Connect your store: https://outercite.com/learn/shopping/connect-your-store (Connect Shopify or WooCommerce, choose the products worth tracking, and understand what a sync does to your caps.) - Shopping overview: https://outercite.com/learn/shopping/overview (Product-level visibility: which of your products AI engines recommend for real shopping questions, and which are invisible. In beta.) - Data handling and privacy: https://outercite.com/learn/trust/data-handling-and-privacy (What Outercite collects, what it is used for, and where to find the formal policy, retention terms and sub-processor information.) - Data isolation: https://outercite.com/learn/trust/data-isolation (How one client's data is kept from another: separation enforced where data is read, not by filtering rows in the interface.) - How we verify citations: https://outercite.com/learn/trust/how-we-verify-citations (The accuracy claim stated precisely: what each model does, when the second one runs, and what happens when they disagree.) - Trust overview: https://outercite.com/learn/trust/overview (Three commitments: the numbers are conservative, the evidence is readable, and one client's data never reaches another.)