Why your business doesn't show up on ChatGPT

This page is a diagnostic, not a pitch. Work down it in order: the reasons are roughly ranked by how often they're the actual problem, and the first one is the actual problem far more often than the industry admits.

1. Your content only exists after JavaScript runs

This is the big one, and it's binary rather than gradual.

According to Vercel's AI crawler study, no major AI crawler renders JavaScript. Not GPTBot, not OAI-SearchBot, not ClaudeBot, not PerplexityBot. They request your URL, read whatever HTML comes back, and leave. PerplexityBot parses static HTML only. GPTBot fetched JavaScript in about 11.5% of requests and executed none.

Googlebot does render JavaScript. This is why the confusing case exists: a site that ranks perfectly well on Google, and is simply not there as far as ChatGPT is concerned. Same site, two different readers, two different results.

The fix: server-render your content, or pre-render it. The test is trivial. Load your page with JavaScript disabled. Whatever remains is what an AI crawler gets. If that's a logo and a loading spinner, you've found your answer.

2. You're blocking the crawlers without knowing it

A surprising number of sites disallow AI crawlers in robots.txt: inherited from a template, added by an agency during a scraping panic, or switched on by a plugin's default. It's a legitimate choice. It's rarely a chosen one.

The fix: read your own /robots.txt and look for GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot, Google-Extended. OpenAI's crawler documentation lists the current user-agent strings. Decide on purpose, then write down why.

3. Nothing on the page says who or where you are, in a form a machine can read

Humans infer your business from the photos, the layout, the phone number in the footer. Machines don't infer; they extract. Structured data (Schema.org, as JSON-LD) is a labelled statement of the facts: this is the name, this is the address, these are the services, these are the hours.

Be careful how you hear this, because the industry oversells it: adding schema does not make an AI recommend you. What it does is remove ambiguity. It's the difference between a machine guessing from prose and reading a label.

I scanned 35 Indian clinic websites and published the results in the AI Visibility Index. Most were fine: the median scored 87/100. The finding was a minority. Four of 35 published no structured data at all, and they hold the three lowest scores in the sample.

The fix: add the schema type that matches what you are, with real values, and don't lie in it.

4. Nobody outside your own website says you exist

Everything above is about your domain. This one isn't, which is why it's the least popular item on the list: there's no snippet to paste.

Answer engines weight corroboration. Claims that appear only on your own site are self-issued; claims that appear on independent pages are evidence. If you're absent from every directory, roundup and local publication in your category, a model has nothing to lean on when deciding whether to name you.

The fix: exist in the sources that get cited for your category. It's slow. It works.

5. The myth: llms.txt

You may have been sold an llms.txt file, or seen a tool generate one. Here's the state of it, with sources, because this page's whole premise is that you deserve the actual evidence:

• Google has said it does not support llms.txt and has no plans to. John Mueller compared it to the keywords meta tag, the one search engines stopped trusting over a decade ago, precisely because the site owner controls it.
• Ahrefs' crawl of ~38,000 domains with a valid llms.txt found that 97% of them received zero AI requests in May 2026. Not "low traffic". Zero.
• No major provider (OpenAI, Anthropic, Google, Meta) has committed to using it as a signal in a production answer surface.

The fix: there's nothing to fix. An llms.txt is harmless, costs ten minutes, and does approximately nothing. If someone quoted you for AI visibility work and llms.txt was a headline deliverable, that tells you something useful about the quote.

The same caution applies to FAQ schema, which gets recommended everywhere. Google's own announcement restricted FAQ rich results to government and health sites in August 2023, and by 2026 they had stopped appearing in search altogether. FAQ markup is still parsed and it's still worth having as clean structure, but the active ingredient is the question-and-answer content itself, not the JSON around it.

The order to work in

1. Disable JavaScript, load your page, see what survives. Fix that first. Nothing else matters if the page is empty.
2. Read your robots.txt. Decide deliberately.
3. Add structured data that matches reality.
4. Start earning mentions you don't control.
5. Ignore llms.txt.

Frequently asked questions

Related