First, find out which cause it is
When search is on, ChatGPT searches the web first, writes its answer from the pages it found, and lists the pages it cites under the answer. The products it recommends come almost entirely from those pages. So when you are left out, it is usually one of the four problems below; check them from the top down:
| Cause | How to confirm | How to fix |
|---|---|---|
| ChatGPT can't reach your site | robots.txt or your CDN blocks OAI-SearchBot | Allow OAI-SearchBot and ChatGPT-User (step 3) |
| It reads an empty page | The text only appears after JavaScript runs | Server-side rendering or prerendering (step 4) |
| You are not on the pages it cites | You are missing from the "best X tools" articles and comparison pages ChatGPT cites | Ask those articles to include you, and write your own comparison pages (step 5) |
| Your page isn't clear | The first screen of your homepage doesn't say what it is, who it's for and what it costs | Write it in sentences that can be quoted as they are (step 6) |
Ranking well on Google doesn't mean you get cited: in Ahrefs' analysis of 15,000 prompts (2025-08), only 8% of the URLs ChatGPT cited ranked in Google's top 10.
Follow these 7 steps
1. Write down 5 questions buyers ask
Leave out your brand name and write them the way buyers would. For example: "AI tool to turn a PDF into editable PowerPoint slides", "AI slide maker without a subscription". If you sell in several languages, write them in each.
2. Ask ChatGPT each question 3 times
Start a new chat each time and make sure search was used (the answer lists sources). Write down three things: whether you were mentioned, who was recommended, and which URLs were cited.
Asking once tells you nothing: in SparkToro's 2026-01 study, asking ChatGPT or Google's AI search the same question again and again returned the same list of recommendations less than 1% of the time. What counts is how many of the 3 answers include you.
3. Make sure OpenAI's crawlers can get in
OpenAI's documentation says sites that block OAI-SearchBot in robots.txt are not shown in ChatGPT search answers; ChatGPT-User is what ChatGPT uses when someone asks it to open a page in a chat. Enter your URL in the AI crawler check and see whether both are allowed. If not, add this to the top of your robots.txt:
User-agent: OAI-SearchBot Allow: / User-agent: ChatGPT-User Allow: /
OpenAI says the change takes about 24 hours to apply. GPTBot is only for model training; blocking it or not makes no difference to ChatGPT search answers. Your CDN's "block AI bots" switch and bot challenges ignore robots.txt, so check those too.
4. Make sure your text is in the HTML itself
Run this line in a terminal, with a sentence from your homepage inside the quotes. If it prints 0, your text only appears after JavaScript runs and crawlers get an empty page:
curl -s https://your-site.com/ | grep -c "a sentence from your homepage"
The fix: render your homepage, pricing page and feature pages on the server or prerender them (Next.js, Nuxt and Astro all can), so curl shows the full text and the prices.
5. Work through the pages ChatGPT cites
Sort the URLs you wrote down in step 2 by type and handle each type:
- "Best X tools" list articles: write the author a short note saying how you differ from the products on the list and what you cost, and ask them to consider adding you.
- "X alternatives" and "X vs Y" pages: write pages on the same topics on your own site, describing the differences honestly.
- Forum and Q&A threads: answer from your own account and say you are the maker.
- Directories: submit your product.
6. Write your homepage in sentences that can be quoted
On the first screen, say in one sentence what you are, who it's for, what it costs and how you differ from the product you are most often compared with. Write features, prices and supported formats as text, not only in images.
Then write an /llms.txt with the llms.txt generator. It costs little, but it is not a rule OpenAI has committed to reading, so do it after the steps above.
7. Test again with the same questions in 2–4 weeks
Use the same questions and the same number of runs as in step 2, and compare how many answers include you, not any single answer.
A real example
Our evaluation of GammaSlide on 2026-10-26: buyers' questions were put to AI (Claude Sonnet 5.5, web search on) 21 times, and 0 answers mentioned it; the most recommended were Gamma, Plus AI, Beautiful.ai. Of the pages the AI cited, 28 were read, 19 listed products in the category, and 0 mentioned it. The homepage's raw HTML had only 314 characters; after rendering in a browser, 18,630.
That round asked Claude, not ChatGPT, but it got stuck exactly where rows 2 and 3 of the table above say: the text waits for scripts, and it is missing from the lists that get cited. Full data in the GammaSlide case study.
FAQ
Will blocking GPTBot stop ChatGPT from recommending me?
No. OpenAI uses separate crawlers for training and search: GPTBot collects pages to train models, OAI-SearchBot serves ChatGPT search. OAI-SearchBot is the one that decides whether you can appear in search answers.
I rank first on Google. Why doesn't ChatGPT mention me?
The pages ChatGPT cites overlap little with Google's top 10: 8% in Ahrefs' 2025-08 analysis. ChatGPT searches several rephrasings of one question, and it often cites list articles and comparison pages rather than the top-ranked page.
How soon will I see a change?
OpenAI says robots.txt changes take about 24 hours. Cited articles have to be updated by their authors and found again, which has no fixed timeline. That is why you test again with the same questions after 2–4 weeks.
Is there a way to guarantee ChatGPT recommends me?
No. The answers differ every time, and nobody can guarantee them. What you can do is make sure ChatGPT can read you and that the pages it cites include you, then count how often you show up.
Sources
- OpenAI · Overview of OpenAI Crawlers （checked 2026-11-03）
- Ahrefs · Only 12% of AI Cited URLs Rank in Google’s Top 10 for the Original Prompt （published 2025-08-11, 15,000 prompts）
- SparkToro × Gumshoe · AIs are highly inconsistent when recommending brands or products （2026-01, 12 prompts run 60–100 times on each platform）
- llmstxt.org · The /llms.txt file （proposed in 2024）
- kagentx evaluations: the data on the case study pages; each example gives its date.