Does ChatGPT read llms.txt?

OpenAI's own crawler documentation does not state that GPTBot, OAI-SearchBot or ChatGPT-User read or parse a site's llms.txt file. OpenAI publishes its own llms.txt for its developer docs, but that is OpenAI acting as a publisher, not evidence that its crawlers consume other sites' llms.txt files.

Does Google use llms.txt?

No. Google's Search Advocate John Mueller said directly in June 2025 that no AI system currently uses llms.txt, and Google's own AI features documentation states that no new machine-readable files or special structured data are needed to appear in AI Overviews or AI Mode.

Does Perplexity read llms.txt?

Perplexity's own crawler documentation does not confirm this. Some SEO vendors claim Perplexity reads llms.txt to improve citation accuracy, but we found no statement from Perplexity itself backing that claim, so it should be treated as unverified marketing, not fact.

What is the difference between llms.txt and robots.txt?

robots.txt is a decades-old, universally respected standard that tells crawlers which URLs they may or may not fetch. llms.txt is a much newer proposal that does not control access at all; it is a curated summary and link list for language models, and no major AI system has documented that it actually reads one.

Should I still create an llms.txt file?

It is still worth an hour for documentation, API and developer product sites, since some coding assistants and agent tools can be pointed at one directly by a developer. It should not be created in the expectation that it will improve citation rates in ChatGPT, Perplexity, Gemini or Google's AI features, since no vendor has documented that its systems read one automatically.

Does Anyone Actually Read llms.txt in 2026?

A lot of AEO advice repeats the same claim: publish an llms.txt file and AI assistants will cite you more. We checked that claim directly against the crawler documentation OpenAI, Anthropic, Perplexity and Google each publish themselves. None of them states that it reads llms.txt. Google has said the opposite, on the record, twice.

What llms.txt is, briefly

llms.txt is a proposed Markdown file at a site's root that gives language models a short summary and a curated link list, so an assistant does not have to crawl a whole site to find the pages that matter. We cover the format and how to build one in what is llms.txt, and should you create one?. This post is narrower: it checks, vendor by vendor, whether any major AI system actually reads the file once it is published.

Google: on the record, twice

Google's Search Advocate John Mueller said directly, in a June 2025 post widely reported at the time: no AI system currently uses llms.txt. Google's own AI features documentation, last updated December 2025, says the same thing in its own words: "You don't need to create new machine readable files, AI text files, or markup to appear in these features. There's also no special schema.org structured data that you need to add." That is a primary Google Search Central docs page, not a social media post, and it rules out llms.txt by name alongside every other AI-specific file format.

OpenAI, Anthropic and Perplexity: no confirming statement either way

Reading each company's own published crawler documentation gives a more precise picture than "nobody uses it":

OpenAI's crawler docs describe GPTBot (training), OAI-SearchBot (search indexing for ChatGPT) and ChatGPT-User (on-demand fetch). None of the three is documented as reading llms.txt. OpenAI does publish its own llms.txt for its developer docs, which is OpenAI acting as a publisher of the format, not evidence that its crawlers consume other sites' files.
Anthropic's support article on its crawlers (ClaudeBot, Claude-User, Claude-SearchBot) covers robots.txt behavior for each in detail and does not mention llms.txt anywhere. Anthropic also publishes its own docs' llms.txt, the same publisher-not-consumer pattern as OpenAI.
Perplexity's crawler documentation mentions llms.txt only as a pointer to Perplexity's own documentation index, the same pattern again. Some SEO vendors claim Perplexity actively reads llms.txt to improve citation accuracy. We found no statement from Perplexity itself backing that claim, so it should be treated as an unverified marketing claim, not a documented fact.

The pattern across all three: a company publishing its own llms.txt file (as a courtesy to developers reading its docs) gets repeated in marketing content as if it were evidence that the company's crawlers also read other sites' llms.txt files. Those are two different claims, and only the first one is actually documented.

Where llms.txt is genuinely used today

The llms.txt spec itself reached version 2 in August 2026, described by its maintainers as incorporating two years of adoption feedback, so the format is not abandoned. Its real, working use case today is narrower than "AI search visibility": a developer using an AI coding assistant can point that assistant directly at a library's llms.txt to get a compact map of its documentation, instead of the assistant crawling the whole site or the developer pasting pages in by hand. That is a genuinely useful, if manual and developer-initiated, workflow, and it is a different thing from an AI search engine automatically discovering and weighting a site's llms.txt when deciding what to cite.

What this means in practice

Do not create llms.txt expecting more AI citations. No vendor has documented that its citation or answer systems read it. That claim, wherever you see it, is unsourced.
Still worth an hour for documentation and developer products. The direct, developer-initiated use case (pointing a coding assistant at your docs) is real and low-cost to support.
Spend the rest of the budget on what is actually documented to matter: crawlable pages, robots.txt rules that match what you actually want (see AI crawlers and robots.txt, explained), and content structured so an answer engine can extract it cleanly, covered in how to structure content so AI can actually extract it.

The distinction worth holding onto

A company publishing its own llms.txt is evidence the format exists and is easy to adopt. It is not evidence that company's crawlers read other sites' llms.txt files. Those are separate claims, and only Google has stated its position on the second one directly.

How we approach this

We still build llms.txt for clients with documentation or developer products, because the coding-assistant use case is real and the cost is small. We do not sell it, or any other single file, as an AI-visibility lever, because none of the vendors whose systems would need to read it have documented that they do.

Sources: Google's AI features documentation, Search Engine Journal on John Mueller's June 2025 statement, OpenAI's crawler documentation, Anthropic's crawler support article, Perplexity's crawler documentation, llmstxt.org.

Does anyone actually read llms.txt in 2026?