Why AI Doesn’t Cite Your Brand: 8 Causes and How to Check Each One

Reading time: 14 minutes

For your brand to appear in AI search, your site must be open to AI crawlers and in the index each engine uses. Your brand must also appear in third-party sources, keep consistent data across channels and have pages that answer customers. When AI search doesn’t cite a brand, one of these conditions usually failed. Here are the 8 causes and how to check them.

Why doesn’t AI cite your brand?

In general, AI search leaves a brand out for one of eight common reasons. They fall into four groups: site access, third-party sources, brand data and content that answers the customer. Each cause has a simple test. The table lists the tests in the order to run them, from technical blocks to category association.

Cause	How to check
robots.txt blocks the AI search crawler	Look for OAI-SearchBot, PerplexityBot and Claude-SearchBot in the file
The CDN or firewall blocks crawlers	Review the AI crawler rules in Cloudflare and the logs
The site is missing from the engine’s index	Inspect the URLs in Bing Webmaster Tools and Search Console
The text depends on JavaScript	Open the source code and look for the main text
The brand is missing from third-party sources	Note the domains cited for category questions
The brand data doesn’t match	Compare what AI says with your site, profiles and directories
No page answers the question	Match each customer question to a page
The brand only shows up by name	Compare questions by name with category questions

The first four stop AI from reading the site and are usually faster to fix. The other four decide whether AI search cites the brand once the site is accessible. They require content and reputation work, which takes longer. To do it all at once, use the free GEO audit checklist.

Can AI access your site?

For AI to cite your site in answers that use search, its crawler must be able to read your pages. Four technical barriers prevent this: robots.txt rules, CDN or firewall blocks, absence from the index and text that needs JavaScript.

robots.txt

Each AI engine has a search crawler, and that’s the one you must allow. For ChatGPT, for example, it’s OAI-SearchBot. According to OpenAI’s documentation, sites that block this crawler don’t appear in answers that use search. However, blocking GPTBot, which collects content to train AI models, doesn’t remove the site from ChatGPT search.

Similarly, Perplexity’s search crawler is PerplexityBot. For Claude, they’re Claude-SearchBot and Claude-User. In the Gemini app, on the other hand, the Google-Extended token controls the use of your content in answers. To check, open your site’s address followed by /robots.txt and look for these names on a Disallow line. The guide on how to show up in ChatGPT covers the difference between OpenAI’s crawlers.

Cloudflare and firewalls

OpenAI also recommends that your host and CDN accept the crawler’s IP addresses. In Cloudflare, the AI policies split crawlers into three types: Search, Agent and Training. Therefore, keep the Search policy allowed.

Since September 15, 2026, choosing Block for AI training also blocks Googlebot, Bingbot and Applebot, search included. So, to refuse training without leaving search, Cloudflare points to the Disallow AI Training option. According to the company, which sells these controls, less than 1% of sites on its network block AI search crawlers. Meanwhile, 17% use some kind of training block.

Sites that already refused training moved to Disallow AI Training, with no effect on search. On other CDNs, check the firewall and bot rules and look for refused visits in the logs.

Bing and Google

The site also needs to be in the index the engine uses. According to Microsoft, Bing powers AI tools like ChatGPT. Check in Bing Webmaster Tools whether your important pages are indexed. See how to index your site in Bing Webmaster Tools.

Google likewise needs the page indexed and eligible for a snippet. Since July 2026, Google’s guide also requires the site to stay included in the Search generative AI control in Search Console. Inclusion is the default, but it’s worth checking for clients with several properties.

JavaScript

Some AI crawlers may read only the HTML the server delivers. In a December 2024 study, for example, Vercel analyzed the traffic of these crawlers. OpenAI’s and Anthropic’s crawlers downloaded JavaScript files but didn’t run them. As a result, for these crawlers, text that only appears after JavaScript runs may stay invisible. Vercel sells hosting for web applications.

The study is almost two years old, though. On September 30, 2026, a developer reported that OAI-SearchBot and GPTBot had started running JavaScript on his sites. Still, OpenAI’s documentation doesn’t address the topic. When in doubt, open the page’s source code and look for the main text, prices and products.

Does your brand exist in the sources AI consults?

Most AI citations come from outside the brand’s site, such as the press, review sites, communities and encyclopedias. A brand missing from these sources therefore gives AI search little material to cite. The test is to see which domains AI search cites in your category and whether your brand is on them.

Muck Rack, which sells public relations software, analyzed more than 25 million links cited by ChatGPT, Claude and Gemini. In its May 2026 report, about 84% of citations came from earned media, such as journalism, research, encyclopedias and social networks. In contrast, owned and paid content added up to about 16%.

To check, run your category questions in two or three engines and note the cited domains. Then look for your brand on each one: articles, comparisons, Reclame Aqui (a consumer complaints site in Brazil), review sites and Wikipedia. Each domain without your brand is a content or PR target. The rules for Reddit, reviews and Wikipedia are in the guide to the sources ChatGPT cites most.

Does your brand information match across site, profiles and directories?

The information matches when name, category, offer, price, address and phone number are the same on every channel. That includes your site, Google Business Profile, social networks, marketplaces and directories. When they differ, AI may describe the brand wrongly or prefer another source. Therefore, the test is to ask AI what it knows about the brand and compare.

Profiles and directories carry weight, especially in local questions. Yext, which sells listings management, analyzed 6.8 million citations from ChatGPT, Gemini and Perplexity in July and August 2025. The questions included a location and involved its clients and prospects. Brands’ own websites accounted for 44% of citations, while listings such as directories and profiles accounted for 42%. In healthcare, listings reached 52.6%, driven by US directories. This sample, across four industries, helps explain why the numbers differ from Muck Rack’s.

How do you check what AI says about your brand?

In an Ahrefs experiment, Mateusz Makosiewicz created a fictional brand with an official FAQ and spread three false stories about it. Five of eight AI products repeated the stories, but only ChatGPT relied on the FAQ. His conclusion:

“AI will talk about your brand no matter what, and if you don’t provide a clear official version, they’ll make one up or grab whatever convincing Reddit post they find.”

Mateusz Makosiewicz, Ahrefs (2025)

The test used one brand, 56 questions and, mostly, the engines’ APIs. It works as an illustration, not as a measurement. To check yours, ask two engines what the brand does, what it costs, where it operates and who founded it. Then note each wrong fact and the cited source. Finally, fix it at the source and on your site, with a visible date. For clinics and stores with an address, check your Google Business Profile first and use the checklist in the guide to GEO for local businesses.

Does your brand content answer the questions customers ask AI?

Your content answers them when each important customer question has a page with the answer right at the start. This matters because AI splits the question into smaller searches, such as price, comparison, use and location. If no page on your site answers these searches, AI search cites whoever does, not your brand, even with an open site.

According to Google Brazil, AI Mode splits the question into subtopics and runs several searches at the same time. Similarly, OpenAI explains that ChatGPT rewrites the request into more specific queries.

For example, “what is the best ERP for manufacturing” can turn into searches about price, plant size and comparisons between systems. A site with only a homepage and product pages doesn’t answer them.

To check, list 20 to 30 real customer questions from sales, support and Search Console. For each one, point to the page that answers it in the first few lines. Questions without a page become content topics. At the same time, Google warns that creating a page for every variation, mainly to manipulate answers, violates its scaled content abuse policy. What the official guide asks for is in how to show up in AI Mode and AI Overviews.

Does AI cite your brand in category questions or only by name?

Many brands show up in AI answers when someone asks for them by name. But they vanish in category questions, such as “best dental clinics in Campinas” (a city in Brazil). In a question by name, AI only needs to find the brand. In a category question, however, it must link the brand to the product type and pick it over competitors. These are, in other words, different problems.

A study by Professor Wilson Silva of ESPM, a business school in Brazil, assessed 552 brands from 22 industries in ChatGPT, Gemini, Perplexity and Claude. According to Mundo do Marketing, a Brazilian marketing outlet, 57% were not cited by any of the four AIs in questions about their own category. According to the article, some brands had high domain authority in Google and zero visibility. Silva is also the founder of WS Labs, which offers a tool derived from the study.

To check, run three types of question:

• by name, such as “what is brand X”;
• category, such as “best ERPs for manufacturing”;
• problem, such as “how to reduce production losses”.

If AI search only cites the brand in the first type, the problem is association. In general, it lacks category pages, comparisons and third-party lists that include it. But if it doesn’t show up even by name, go back to the access and source causes.

Why doesn’t AI cite a new brand?

New brands face one more obstacle. After all, without search, the AI model answers with what it learned in training, and the brand isn’t there. The guide to what GEO is explains this difference between AI memory and search.

What doesn’t explain the absence: llms.txt, schema and plugins

Missing llms.txt, schema or a GEO plugin doesn’t explain why AI search doesn’t cite a brand. In fact, Google says Search ignores text files made for AI. Also, the available studies show no citation gain from llms.txt or schema. Fixing the eight causes above therefore comes first.

In an Ahrefs study of 137,000 domains, 97% of llms.txt files received no requests in May 2026. The file’s usage data is in the guide on what llms.txt is for.

In another Ahrefs test, 1,885 pages that added schema saw no meaningful citation gain compared with 4,000 control pages. Ahrefs sells SEO and AI monitoring tools. A plugin can generate these files and markup, but that doesn’t fix any of the eight causes. See what GEO plugins for WordPress do, and what they don’t.

How do you measure whether the fix worked?

A fix has worked when the brand’s presence rises across a fixed set of questions, measured before and after the change. Google’s and Bing’s reports should also show more impressions and citations. A single answer proves nothing, because the list of brands changes with every run.

In research by SparkToro with Gumshoe.ai, which sells AI monitoring, 600 volunteers made 2,961 runs of 12 questions. They used ChatGPT, Claude and Google’s AI. The chance of ChatGPT repeating the same list of brands in two answers was below 1 in 100. So run each question several times and compare the presence rate, not a single answer.

Each fix has its own timeline. According to OpenAI, a robots.txt change takes about 24 hours to take effect in ChatGPT search. According to the IndexNow FAQ, page discovery by crawlers can take days or weeks. On the other hand, the effect of new pages and of mentions on other sites has no set timeline.

Which reports show your brand in AI search?

Search Console’s Generative AI performance report shows the site’s impressions in AI Overviews and AI Mode. Similarly, the AI Performance report in Bing Webmaster Tools shows citations by page in Copilot and select partners. In the logs, the return of OAI-SearchBot shows that the block no longer applies. The full protocol is in the guide on how to measure your brand’s mentions and citations.

What we found when we analyzed 11 pages in Portuguese on the topic

We ran 3 searches in Portuguese about how to get cited by AI and why a company doesn’t show up in ChatGPT. Then we analyzed 11 pages in Portuguese that appeared in them. Five point to inconsistent data across channels as a cause. However, only three cover crawler blocking, and only two name OAI-SearchBot.

Two link ChatGPT to Bing, but none mentions Bing Webmaster Tools or IndexNow. Only one mentions JavaScript, and none covers Cloudflare’s September 2026 change. Also, five recommend schema to get cited, but none mentions the test in which it brought no meaningful gain. Finally, none says how many times to repeat each question or mentions the AI reports in Search Console or Bing.

Frequently asked questions

Main sources

• Cloudflare, Have it both ways: stay discoverable in search while disallowing AI training (2026).
• OpenAI, Overview of OpenAI Crawlers (2026).
• Google Search Central, Google’s Guide to Optimizing for Generative AI Features on Google Search (2026).
• Ahrefs, I Ran an AI Misinformation Experiment. Every Marketer Should See the Results (2025).
• Mundo do Marketing, ESPM study of 552 brands, in Portuguese (2026).
• SparkToro, research on how inconsistent AI brand recommendations are (2026).