Why your brand may be missing from ChatGPT, from crawl access and unclear positioning to weak outside sources and incomplete testing.
Why is my brand not in ChatGPT? Usually, the problem is not a single setting. ChatGPT may be unable to read your site, may not understand what you sell, may lack useful outside pages, or may simply not have tested the buyer question that matters to you.
The rest of this post covers the five causes we see in scans, then gives you a check you can run in under an hour.
A business cannot be selected from a page an engine cannot fetch or understand. In CiteGraph's readiness checks, AI crawlers were allowed in robots.txt on 116 of 121 sites, so access was not the main problem for most of the sample. Five sites failed that check. Content was present without JavaScript on 92 sites, failed on 18 and warned on 9. A page can look complete in your browser while offering much less usable text to a crawler.
The first job is to see what a crawler can actually reach.
Other technical signals help an engine join the dots. Organization schema passed on 72 of 121 sites and failed on 47. Product or SoftwareApplication schema passed on 60 and failed on 59. Canonical and Open Graph tags passed on 92, with warnings on 27. These checks do not make a business eligible by themselves, but they reduce the number of ways a business can be hard to identify or associate with a page.
Use the free AI readiness check, then inspect the rendered page, robots.txt and important product pages.
A sitemap is useful, but it cannot rescue a page that says very little. The readiness sample had a published sitemap on 100 of 121 sites, while the homepage answered a plain request on 119. The practical fix is to make the important facts available in the page text, rather than keeping them inside an interface.
A homepage can be polished and still leave an engine guessing. The important sentence needs to name the thing, the audience, the price or pricing model, and the alternative it replaces or sits beside. In CiteGraph's readiness sample, only 20 of 121 sites passed the check for that plain statement. Forty-three failed and 56 warned. This was among the weakest checks in the set.
The missing sentence is often more important than another design pass.
Metadata helps, but it is not the whole explanation. Title and description said what the site was on 92 of 121 sites, with warnings on 27. One clear H1 passed on 101 sites and failed on 18. Those results are healthier than the plain business statement, which suggests that many sites have the usual page furniture without a sentence that gives an engine a reliable answer.
Write the sentence as if a buyer had never heard of you: We are a category for a customer that helps with a job, from a price or pricing basis, instead of an alternative.
Treat that as a test rather than finished copy. A stranger should be able to repeat the facts without opening five tabs. This is a positioning problem before it is a wording problem.
Your own site is one possible source, and it is not the source type that dominates CiteGraph's citation data. Across 45,309 citations in 7,260 AI answers, listicles and roundups accounted for 67.5 percent. Competitor product pages accounted for 24.4 percent. Together, those two page types made up 91.9 percent of citations. The product's own site accounted for 1.5 percent.
That means the fastest way to be seen is rarely your own homepage. It is often getting into the pages already being cited.
Our public research page records:
“the product's own site take 1.5% of all citations we have recorded”
This is a useful warning, not a rule that your homepage does not matter. The figure describes CiteGraph's recorded sample, where listicles, roundups and competitor product pages dominate. Your task is to find the outside pages that answer your buyers' questions and check whether they describe your business accurately.
Outside context carries more weight in this sample than a company's own pages.
That does not mean you should try to force your business into a roundup. It means the answer often depends on pages that compare, shortlist or describe products. Community threads accounted for 3.4 percent, documentation for 1.1 percent, review platforms for 0.8 percent and social posts for 0.8 percent. News accounted for 0.1 percent, while Wikipedia accounted for 0 percent in the recorded share, despite 21 citations.
Find the pages that already answer your buyers' questions, then check whether they name you accurately. Comparison pages, category roundups and competitor alternatives are sensible places to inspect because they are the page types most often cited in this sample.
A business can be clear and well covered, yet absent from an answer because the prompt does not match the job it does. CiteGraph's scans asked 442 distinct buyer questions across 44 categories. They were not a complete record of every way a buyer could describe a need, and they were not a guarantee that every product would be tested against its best use case.
A missed prompt may tell you more about the test than the business.
This matters when you check the result with a vague prompt. A request such as software gives an engine little reason to retrieve a specialist product. A buyer-shaped question gives it more to work with: a category, a use case, a constraint, an audience or an alternative. The readiness check for comparison or alternatives pages passed on 54 of 121 sites, failed on 21 and warned on 46. Questions answered on the page passed on 56, with warnings on 63.
The translation is simple: test the job your buyer is trying to complete, not a broad label for your market.
Build or improve pages around the questions your buyers already ask in sales calls, support tickets and comparison conversations. Answer the question directly, then explain who the product suits, who it does not suit, what it costs and what it replaces. Create pages where the buyer's task and your product's evidence meet.
List questions recent buyers actually asked, then test those first.
An absent answer is weak evidence. AI answers vary by prompt, time, engine, retrieved pages and the sample used to test them. CiteGraph's recorded data came from 60 scans of 41 sites, covering 44 categories. Those scans produced 45,309 citations from 7,260 answers, drawing on 7,624 domains and 16,506 pages. That is useful evidence, not a census of every answer an engine could give.
A missed answer is a clue, not a verdict.
A scan can miss a real mention. It can use a different interpretation of the category, select a different comparison set or find a page you did not expect. The reverse is also possible: a favourable answer does not establish steady visibility. Measure repeated buyer questions and inspect the cited page, rather than treating a prompt as a verdict on the business.
We measure samples, so we do not know what we have not measured. The methodology page explains how CiteGraph measures these results.
Use the free AI visibility checker to test the exact questions that matter to your buyers. Keep the prompt, date and answer together. The weekly research on who owns AI answers shows the wider pattern.
The question to ask next is: which of these five problems is blocking your business first? The answer lives in the page evidence, the outside sources and repeated scans, not in a single ChatGPT response.
A page can look complete in a browser while offering less usable text to a crawler. Check robots.txt, the rendered page and important product pages, because five of 121 sites failed the crawler access check and content failed without JavaScript on 18.
Only 20 of 121 sites passed the check for clearly stating the business. Explain the category, audience, price or pricing model, and alternative it replaces or sits beside.
Listicles and roundups accounted for 67.5 percent of recorded citations, while competitor product pages accounted for 24.4 percent. Inspect those pages to see whether they name your business accurately, because the business's own site accounted for 1.5 percent.
A prompt may not match the job your product does, and an absent answer is weak evidence because results vary by prompt, time, engine, retrieved pages and sample. Test several buyer-shaped questions that include a use case, constraint, audience or alternative.
Cite this: CiteGraph, “Why is my business not in ChatGPT? Five reasons from scans”, 11 Sept 2026, https://www.citegraph.app/blog/why-is-my-business-not-in-chatgpt