Why doesn't ChatGPT recommend my business?
Usually one of four fixable problems: a blocked crawler, a JavaScript wall, content that is too thin to extract, or too little proof of you outside your own site. Here is how to tell which one and fix it.
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, and for the ChatGPT-specific do-this-now checklist, see 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). 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). 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). 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). 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).
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). 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; 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.
- Check robots.txt at yourdomain.com/robots.txt for a Disallow rule against OAI-SearchBot or user-agent *.
- 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.
- Read your homepage as a stranger would: does the first paragraph state plainly what you do, for whom, and where, with specifics?
- Search your business name plus your category. If your own website is the only result, you have no third-party presence yet.
- 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.
- GPTBot blocked by 5.89% of roughly 140 million websites (7.3% including subdomains), ClaudeBot block-rate growth: Ahrefs, 2025. Study.
- GPTBot fetching but not executing JavaScript on roughly 11.5% of requests: Vercel and Merj, 2024. Analysis.
- 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.
- The 85% earned-media citation share across more than one million AI prompts: Muck Rack, 2026. Study.
