A ground-up answer engine audit. We query six AI engines live, verify every citation with two models, then grade the four levels of your site that decide whether you get named at all.
We use this to ask AI engines the same question your customers ask.
Free · No account required · 3 checks per hour
Answer engine optimization is a stack, and it fails bottom-up. A blocked crawler makes perfect schema irrelevant; unreadable content makes authority signals moot. We grade every level so you know which one is actually costing you.
Can AI engines reach and read your site?
We separate the crawlers that fetch pages in order to cite them, OAI-SearchBot, PerplexityBot, Claude-User, Googlebot, from the ones that only gather training data. Blocking the first group costs you answers. Blocking the second is a licensing choice we will not mark you down for, whatever other tools tell you.
Can AI parse and extract your content?
AI systems extract passages, not pages. We grade your title, meta description, heading hierarchy, canonical URL, entity schema, and, weighted highest, whether your content survives without JavaScript, since no major AI crawler except Gemini and Applebot executes it.
Does AI have reason to trust you?
Trust is the tiebreaker between similar businesses. We look for sameAs links tying your entity to its profiles, machine-readable contact details and address, review markup, freshness dates, and an about page that says who is behind the business.
Is AI recommending you today?
We ask six AI engines a generic question in your category and location. Your name is never in the prompt, then run every answer through a judge model and an independent verifier. Only citations both models confirm are counted.
Most visibility tools string-match your brand name in an AI response and call it a citation. That is how you end up celebrating a mention of a different company with a similar name.
Each engine gets a generic "find me a [service] in [location]" prompt. Your business name never appears, so nothing is led.
A high-performance model reads the full response and decides whether your business was genuinely recommended, with reasoning.
A second, independent model audits that judgement. Disagreement is recorded rather than resolved in your favour.
Only citations both models confirm count, and each carries a confidence score. The target is zero false positives.
An AEO (Answer Engine Optimization) audit measures whether AI assistants can reach, understand, trust, and ultimately recommend your business. Unlike a traditional SEO audit, which grades how you rank in a list of links, an AEO audit grades how citable you are inside a generated answer.
SEO tools optimise for ranking positions. This tool queries the AI engines directly to see whether you are named in their answers, then traces that outcome back to the site-level signals that cause it, schema markup, crawler access, entity links, and server-rendered content.
ChatGPT, Claude, Gemini, Perplexity, Grok, and DeepSeek. Each is queried live with a generic "find me a [service] in [location]" prompt that never mentions your business name, so a citation is earned rather than prompted.
Every response goes through a two-model pipeline: one model judges whether the business was genuinely cited, and a second, independent model audits that judgement. Only findings both models agree on are reported as citations, with a confidence score attached.
Because the evidence says it should not. Ahrefs analysed server logs across 137,210 domains and found 97% of published llms.txt files were never requested by any bot, and that no AI crawler ever went looking for one that was missing. SE Ranking modelled 300,000 domains and found removing the variable made their citation predictions more accurate. Google states plainly that Search ignores it. We report whether you have one, because Lighthouse now audits for it, but we will not mark you down for skipping a file the engines do not read.
Not on the available evidence, and we would rather say so. Ahrefs ran a matched difference-in-differences study over 1,885 pages that added schema against 4,000 controls: no significant gain for ChatGPT or AI Mode, and a significant loss for AI Overviews. Schema is still worth having. It is the cheapest way to tell engines which real-world entity you are, which matters when several businesses share a name, so we score it for entity clarity rather than pretending it lifts citations.
No, but without it only Level 4 (live citations) can run. With a URL we also audit access, structure, and authority, which is where the fixable causes of poor AI visibility almost always live.
The audit is, genuinely: no account, no credit card, three checks per hour per visitor. What it gives you is a one-time snapshot. Tracking the same engines continuously is the paid product, which starts with a 7-day trial.
This audit is a point-in-time read. Outercite re-runs it on a schedule, watches your competitors, and tells you which move to make next.