Does ChatGPT use llms.txt?
The honest answer in 2026: no major AI provider has confirmed reading llms.txt to answer questions. Here is what each has said, and where the file is genuinely read.
Last updated:
The short answer
As of 2026, no major AI provider has officially confirmed that its assistant reads llms.txt at inference time. That includes OpenAI (ChatGPT), Anthropic (Claude), and Perplexity, all of which are sometimes assumed to use it.
This does not make the file useless. It means a defensible use case today is a client or retrieval workflow that is explicitly configured to consume it, not an assumed behavior of a consumer chatbot.
OpenAI and ChatGPT
OpenAI has not published anything stating that ChatGPT, or its GPTBot and OAI-SearchBot crawlers, parse llms.txt or treat it as a special input. When ChatGPT browses the web to answer a question, it retrieves and reads pages; there is no public commitment that it preferentially loads your llms.txt first or uses it as a site map.
OpenAI does publish a llms.txt file for its own developer documentation. That proves publication, not that ChatGPT Search or an OpenAI crawler consumes other sites' files as a special signal.
Fetching is not the same as using
Site owners sometimes see AI crawlers requesting /llms.txt and /llms-full.txt
in their server logs, including bots from OpenAI and Microsoft. It is tempting to read that as proof
the file is being used. It is not.
Anthropic and Perplexity
Anthropic publishes llms.txt and llms-full.txt files for its own documentation. This does not establish how Claude or Perplexity handles files on other sites. Neither provider has publicly documented automatic inference-time use of llms.txt.
Anthropic does operate documented crawlers (for example ClaudeBot for training data and a separate user-triggered fetcher), which respect robots.txt like any well-behaved bot. That governs access, not whether llms.txt is treated as a curated signal.
Where llms.txt does get read
The controllable use case today is the developer-tooling and retrieval layer:
- Agent coding tools can be configured to fetch llms.txt as project or library context before helping a developer.
- RAG pipelines built in-house often use llms.txt as a curated reading list to decide which pages to ingest.
- MCP integrations and documentation platforms can expose the file as part of an explicit workflow.
These uses are testable in the workflow you control. They do not imply better answers from an unconfigured third-party assistant.
What to do today
- Consider publishing llms.txt if you run docs, an API, a developer tool, or technical SaaS and have a compatible client or ingestion workflow to test.
- Do not expect consumer chatbots to read it yet, and do not market it as if they do.
- Measure what you can: server-log hits from AI bots, referrals from AI products, and manual spot-checks of how assistants describe your product. See does llms.txt work? for the full method.
FAQ
Does ChatGPT read llms.txt? OpenAI has not announced that ChatGPT or its crawlers parse it. Fetching the file is not the same as using it.
Does Claude or Perplexity use llms.txt? Not confirmed at inference time, even though both publish their own llms.txt for documentation.
So why publish it? Because a compatible agent, integration, or RAG pipeline can be configured to use it as a curated reading list, and that behavior can be tested directly.
Next steps
- Does Google use llms.txt?
- AI crawlers explained: GPTBot, ClaudeBot, PerplexityBot.
- How to create an llms.txt file.
- Generator: build one from your sitemap.