Skip to content
Tomas KadlecBook a Call

Articles · 18 November 2026 · 3 min read

My brand showed up in 0 of 60 AI answers

I asked an AI with web search 60 buyer questions in 5 languages. My brand was in none. How I measured it, and why the API and ChatGPT disagree.

I asked an AI with web search 60 questions that my buyers ask. In 5 languages. My brand was in 0 answers. My site was cited 0 times. It was not even among the pages the AI read.

That is my baseline, measured on 16 November 2026, before any work on it.

I sell AI visibility work. So yes, this number is a bit uncomfortable to publish. But you can not show progress without a starting point, and most brands never write theirs down.

How I measured it

  • 12 questions a real buyer asks, with no brand names. From "how do I find out if AI recommends my business" to "how often should I check it".
  • 4 buying stages: awareness, engagement, conversion, loyalty. 3 questions each.
  • 5 languages: English, Czech, German, French, Spanish. Search location set per language.
  • 1 engine for breadth: OpenAI's API with web search, so every chat runs the same way and is logged.
  • 3 things logged per chat: was the brand in the answer, which URLs were cited, which URLs the search read.

Result: search ran in 60 of 60 chats. 304 different cited URLs, 1,848 different pages read. My brand: 0, 0 and 0.

The surprise: who gets cited

The API cited mostly official documentation. Google developer pages in 41 of 60 chats, research papers on arxiv in 32, OpenAI developer pages in 31. Same picture in all 5 languages and all 4 stages.

You can not get into Google's own documentation. So if I planned only from this list, I would plan nothing.

Then I asked one of the same questions in normal ChatGPT, in a temporary chat. It cited something completely different: blogs and help centres of AI visibility tools. No official documentation at all.

One chat is not data. But it shows the trap: the API and the app your buyers use do not give the same source map.

How to set up your own baseline

  1. Write 10 to 12 buyer questions for one product. No brand names. Spread them over the buying stages.
  2. Run each in every language you sell in. Answers change a lot between languages.
  3. Log three columns: brand mentioned yes or no, cited URLs, and pages read if your tool shows them.
  4. Check 5 of the questions in the real app (ChatGPT, Gemini, Perplexity), the one your buyers use. Compare the sources with the API list.
  5. Save it with the date. Then measure again after each change, with the same questions.

Here is the prompt I use to sort the cited sources:

I attach a list of URLs that AI assistants cited when answering questions about my topic. Group them by domain. For each domain say which type it is: official documentation, big media, vendor blog, tool directory, forum, or independent blog. Mark which ones a small company can realistically get into (an article, a listing, a guest post, a correction) and which ones not. Do not guess numbers I did not give you. Return one table sorted by the number of citations.

What I do next

I will not try to get into Google's documentation. I will look for the questions where the real app cites blogs, lists and tools, because those pages can change. Then I measure the same 60 questions again and publish the new number, also if it is still 0.

Before any of that, check the basics: can AI crawlers read your pages at all? If your site blocks them, no content work helps.

Summary

I measured my own AI visibility before doing any work on it: 60 chats, 5 languages, 0 mentions, 0 citations. The API mostly cited official documentation, while normal ChatGPT cited vendor blogs. So measure in the app your buyers use, not only in the API, write the baseline down with a date, and measure again with the same questions.

Start with the free AI crawler check, it shows in a minute if AI crawlers can read your site. If you want the full picture for your brand, with questions in your languages and the source map, that is the AI-visibility audit. Or book a free 20-minute call.

Have you ever written down your AI visibility baseline?

Sources: Own measurement on 16 November 2026: 12 questions × 5 languages = 60 chats, OpenAI API with web search, plus one check in ChatGPT (temporary chat). Raw data on request.

Want this in your team?

I set up AI workflows for marketing and sales teams, with the checks built in. Book a free 20-minute call.

Book a free 20-minute call