Why doesn't my brand show up in ChatGPT? Three things to check

If you ask ChatGPT about your brand or your service and you are not in the answer, the place to start looking is the part you can control. This article goes through those parts in order and shows, for each, what we measured and what we changed on our own site.

First, check which question you are missing from

"We don't show up in ChatGPT" can mean two different problems. One: when asked for by name, the engine does not recognise you, or gets you wrong. Two: when asked about your service — "Can you recommend an SEO agency in İzmir?" — your name never appears in the answer. The first is an identity problem, the second a visibility problem, and they are fixed differently.

Our own measurement shows the split clearly. We could not measure ChatGPT: it requires an account, and we do not log in to any AI service. We measured on Perplexity, which works without one. In the first measurement of our October panel (23 September 2026) we were named in all four brand questions and in none of the twelve non-brand questions we could measure. Asked for by name, the facts come back correctly — with one exception, below; asked about our service, we are not there.

1. Can ChatGPT's crawler reach your site?

OpenAI sends more than one crawler to websites, and each does a different job. The table below is based on OpenAI's own crawler documentation.

The subtle point: according to OpenAI these settings are independent of each other. You can block GPTBot and allow OAI-SearchBot; your content is then not used for training, but you can still appear in search. The opposite mistake is also possible: a robots.txt line that blocks AI bots wholesale also blocks OAI-SearchBot. OpenAI writes that a robots.txt change takes about 24 hours to reach its systems.

robots.txt is not the only door. Some hosting and firewall services offer a single setting that blocks AI bots; Vercel's firewall, for example, has a ready-made "AI Bots" ruleset for this. Such a setting works independently of robots.txt: the server can still turn a bot away while robots.txt allows everything.

We measured this on our own site on 24 September 2026. Our robots.txt allows every crawler. Requests sent as OAI-SearchBot, ChatGPT-User, GPTBot, PerplexityBot, Claude-SearchBot, ClaudeBot, Google-Extended and Bingbot to the home page, the llms.txt file and one article all came back successfully (HTTP 200). The test has a limit: the requests come from our own connection and only imitate the crawler's identity. Because a firewall can also recognise real bots by IP address, the definitive answer is the setting in the hosting dashboard.

After robots.txt and the firewall, the third thing to check is the page's text. According to a measurement Vercel and MERJ published in December 2024, the crawlers of OpenAI, Anthropic and Perplexity did not execute JavaScript (OpenAI's and Anthropic's crawlers downloaded JavaScript files but did not run them); Google's Gemini, which uses Googlebot's infrastructure, could. So a page whose text is built later in the browser by JavaScript can show up in Google and still reach ChatGPT's crawler incomplete or empty. On our site the text is generated on the server and arrives in the first HTML; the checks we run before a release read the page without executing JavaScript to confirm it.

• Open robots.txt: is there a "Disallow: /" under "User-agent: OAI-SearchBot" or "User-agent: *"?
• Is a setting such as "AI bots" or "bot protection" switched on in your hosting or CDN dashboard?
• Is the page's main text in the first HTML response? If you can read it with "view page source", yes.

2. Is your brand one consistent entity?

The logic is simple: the engine reads about you from different sources. If those sources do not carry the same name, address, phone number and description — or another company carries the same name — it becomes harder for it to build one correct answer.

That is exactly what happened to us. In the same measurement, when we asked for our contact details the engine gave our address, phone and e-mail correctly — but stopped to ask the user whether they meant a similarly named company in Germany. In response we added a sentence describing who we are to our site's structured data and to our llms.txt file: our domain, our district and our city. We did not name the other company. Whether it worked will show in the November panel (measured at the end of October), and we will add the result to this article.

The sentence in our structured data reads: "Maora Studio (www.maora.studio) is a digital agency based in Karabağlar, İzmir, Türkiye; it is not the same organisation as similarly named companies outside Türkiye." It is short, carries three facts that can be checked (domain, district, city), and does not contradict anything else on the site. Use the same test for your own sentence: pick details that appear the same way on your Google Business Profile and in directories.

• Are your name, address and phone identical on your site, your Google Business Profile and in directories?
• Does your site's structured data (Organization, LocalBusiness) say the same thing as the page?
• Do your social profiles link to your site, and your site to them (sameAs)?
• Is there another company with the same or a similar name? If so, write a sentence that sets you apart by location and domain; do not name the other company.

3. Do you have a page that answers the question?

Solving the brand question does not solve the service question. When someone asks "Who builds booking websites for tour and transfer companies in İzmir?", the material the engine works with is pages that answer that question; in our October record, too, the answers to service questions were built from directories, competitors' own sites and publications. Without such a page, it has less to name you with.

In the 23 September measurement we asked exactly that question, and the engine misread it: it listed transfer companies rather than agencies that build sites. That day we also had no page answering the question. Now we do: a page on booking websites for tour and transfer companies, with the details of two real projects. We will measure the result in the November panel too. The rule we follow on these pages is simple: the question in the heading, the answer in the first sentence, every paragraph able to stand on its own.

Does the engine hear about you anywhere else?

Our own record shows that for service questions the engine does not build the answer from your site alone. Of the twelve non-brand questions we could record in the first measurement, three cited a directory, three cited competitors' own sites and two cited foreign publications; for the remaining four we did not record the source type. For brand questions the source was our own site every time. As far as we know, no independent source mentions us today; that is why directory listings and customer reviews are our next job.

By independent sources we mean:

• Customer reviews on your Google Business Profile.
• Listings in city and industry directories, with details identical to your site's.
• Member pages of the associations or chambers you belong to.
• Local news, interviews and trade publications that mention you.

A "designed by …" link on your clients' sites is a signal too, but it is not independent: you put it there. When you ask for reviews, follow Google's review policy: do not ask only satisfied customers, do not discourage negative reviews, do not ask for specific content, and offer no discount, gift or payment in return.

The source type also tells you what to work on. If the answer is built from a directory, the first step is to be listed there with correct details; if from competitors' sites, to make sure your own page answers the same question plainly; if from foreign publications, to publish a citable article on the subject in your market's language. In the same record the answer to the GEO question was built from ten foreign sources. Our GEO guide and our SEO, AEO and GEO comparison are aimed at exactly that gap: citable, and in Turkish as well as English.

How do you check for yourself?

1. 1.Write the questions: two or three brand questions (with your name) and five or six service questions (without it).
2. 2.Ask ChatGPT from your own account and Perplexity without one; keep a dated screenshot of every answer.
3. 3.For each answer note three things: were you named, in what position, and which source was shown.
4. 4.Do not leave an engine you could not ask blank; record it as "not measured", with the reason. Otherwise you cannot compare two months.
5. 5.A month later ask the same questions in the same order. A one-off check says almost nothing, because answers vary.

Every month we record the same twenty questions for six engines: we ask them where we can reach the engine, and record the rest as "not measured", with the reason. In the first measurement, on 23 September, we could measure 16 of the 120 cells; 100 could not be measured because they required an account or hit Google's bot check, and 4 because Perplexity's limit for signed-out queries ran out. It looks like little, but because it will be measured the same way every month, it will show the change.

What not to do

• Paying for services that guarantee a ranking or a listing in ChatGPT. The engine writes the answer; nobody can guarantee it.
• Blocking every AI bot with one robots.txt line, then wondering why you do not appear in search.
• Producing pages that answer no question and only carry keywords.
• Looking once and deciding.