Should I still publish llms.txt?

Yes, if you can keep it short, accurate, and aligned with canonical pages. It is low cost, but it should not replace Bing indexing, crawlable HTML, robots.txt access, and useful source pages.

Question

Does ChatGPT use llms.txt?

Q: Does ChatGPT use llms.txt?

There is no public confirmation that ChatGPT or any major AI engine uses llms.txt as a retrieval or ranking input. Treat llms.txt as a curated source map that may help machines understand your preferred pages, not as a guaranteed ChatGPT citation channel.

Updated June 8, 2026

There is no public confirmation that ChatGPT, OpenAI search, or any other major AI engine uses llms.txt as a retrieval input, ranking signal, or citation requirement. Treat the file as a curated source map that may help machines understand which public pages you consider canonical, not as a guaranteed ChatGPT citation channel.

What ChatGPT search can use

For practical site owners, the safer assumption is that ChatGPT visibility depends on ordinary discovery and retrieval mechanics first:

Search index discovery: pages that are indexed and can be found for the query.
OAI-SearchBot access: OpenAI's search crawler must not be blocked by robots.txt, Cloudflare, WAF rules, or origin security.
ChatGPT-User fetches: live user-triggered retrieval needs the page to return a clean 200 without login, challenge pages, or broken TLS.
Extractable answers: the page needs direct, quotable text rather than vague marketing copy or JavaScript-only content.

So is llms.txt useless?

No. A good llms.txt can still be useful because it is a concise editorial map of your best source pages. It can tell a crawler or assistant which URLs are worth reading first, what the site is about, and which pages are canonical. That is valuable only when the linked pages themselves are strong.

Where llms.txt helps

Use case	Useful?	Why
Listing canonical documentation pages	Yes	It reduces ambiguity when a site has many archives, tag pages, or duplicate docs.
Getting a page indexed by Google or Bing	No	Use XML sitemaps, internal links, Search Console, Bing Webmaster Tools, and IndexNow.
Guaranteeing ChatGPT citations	No	No public standard says ChatGPT must cite pages listed in `llms.txt`.
Helping assistants choose your best sources	Possibly	The file can be read cheaply and points to your preferred pages, but support varies.

How to publish it without overclaiming

Put the file at https://example.com/llms.txt.
List only stable canonical URLs that return 200.
Keep descriptions factual and short.
Update it when important pages change, but avoid daily churn for no reason.
Link to it from your footer or developer resources if it is part of your public AI policy.

What to verify in logs

Do not judge success by whether the file exists. Check whether search and AI retrieval agents can actually access the pages you care about:

grep -Ei 'OAI-SearchBot|ChatGPT-User|Bingbot|Googlebot' /var/log/nginx/access.log \
  | awk '{print $7, $9, $0}' \
  | tail -50

If your pages are not indexed or not crawled, fix that before polishing the llms.txt wording. The file is a supporting signal, not a substitute for crawlable pages.

Generate your llms.txt