Does ChatGPT read llms.txt?

There is no confirmation from OpenAI that llms.txt influences ChatGPT's answers, and large-scale log analysis shows AI bots do not request the file on sites that lack one. GPTBot occasionally downloads existing files without any observed effect on citations.

Should I still create an llms.txt file?

If it takes minutes and you understand it changes nothing measurable for AI search, it is harmless. Create one with real expected value only if you publish developer documentation that AI coding assistants consume.

Is llms.txt the same as robots.txt for AI?

No. robots.txt controls crawler access and is honoured by major AI crawlers; llms.txt merely offers a content summary that engines are not committed to reading. Access control happens in robots.txt and your CDN rules.

What should I do instead of llms.txt for AI visibility?

Verify AI crawlers are not blocked at your CDN or firewall, restructure key pages to answer first, add sourced statistics, and build mentions on third-party sites AI engines cite. Each has measured effects; llms.txt does not.

llms.txt: What the Data Says Before You Add One

An llms.txt file will almost certainly not improve your AI search visibility: in an Ahrefs analysis of 137,000 domains, 97% of published llms.txt files received zero fetches in a full month, 96% of the few requests that did arrive came from bots rather than answer engines, and no major AI company — OpenAI, Google, Anthropic, or Meta — has committed to reading the file in production answering. Google's official documentation lists it among the things you do not need. The one evidence-backed exception: documentation for developer tools, which AI coding assistants genuinely fetch.

TL;DR

97% of llms.txt files received zero fetches in a month (Ahrefs, 137K domains).
No major AI engine has committed to reading the file; Google lists it as unnecessary.
The one real use: documentation for AI coding tools — Claude Code was the #2 named fetcher.

What llms.txt was supposed to be

Proposed in late 2024, llms.txt is a Markdown file at your domain root that summarizes your site and links its key pages in a form convenient for language models — a curated map, analogous in spirit to robots.txt or sitemaps. The proposal is reasonable; the adoption it assumed never arrived on the answering side.

The numbers, specifically

The Ahrefs study is the largest public dataset on the question. Of 137,000 domains analyzed, about 28% had published an llms.txt file. Ninety-seven percent of those files got no traffic at all in the measurement month. Of the fetches that did occur, 96% were bots, and a meaningful slice came from the industry studying itself — GEO tools, llms.txt checkers, researchers. Most telling: AI bots issued zero requests for llms.txt on sites that did not have one. Crawlers are not looking for it; they simply download it occasionally when it happens to exist. Separate tracking by Search Engine Land across ten sites reached the same conclusion: no measurable visibility effect.

The one place it earns its keep

AI coding tools read documentation manifests. In the Ahrefs fetch data, Claude Code was the second most active named fetcher of llms.txt, and tools like Cursor use such files to pull clean docs into context. If you ship a developer tool, an llms.txt (and llms-full.txt) pointing at your docs helps real users in real workflows. That is a developer-experience feature, not a search-visibility tactic.

The honest cost-benefit

Generating the file takes minutes and harms nothing, so "just in case" is a defensible position — as long as nobody bills you for it or reports it as AI optimization work. The failure mode worth avoiding is substitution: teams shipping an llms.txt instead of fixing crawler access at the CDN (which silently breaks visibility on roughly a quarter of sites), restructuring pages answer-first, or publishing anything worth citing. If a proposal leads with llms.txt, ask what evidence supports the rest of it.