No. Google has stated that llms.txt files are not needed for Google Search, and that keeping one will neither harm nor help your visibility or rankings. An llms.txt file is a community proposal for handing AI agents a tidy markdown index of your content. It can be genuinely useful for documentation, but it is not a ranking signal, and no search engine requires it.
Key takeaways
- Google says llms.txt is not needed for Google Search and has no effect, positive or negative, on rankings.
- The file is a 2024 proposal: a markdown index an agent reads on demand, not a directive crawlers obey.
- OpenAI, Anthropic and Perplexity document robots.txt to control their crawlers, not llms.txt.
- Publish one if you have documentation worth indexing. It is never the fix for a site AI systems cannot fetch or read.
What is an llms.txt file?
llms.txt is a proposal published at llmstxt.org by Jeremy Howard in September 2024 and still maintained there. The reasoning is practical rather than promotional: an HTML page wraps its information in navigation, ads and JavaScript, converting it back into clean text is imprecise, context windows are still too small for a whole website, and, as the proposal puts it, every wasted token costs time and money. So instead of making an assistant reconstruct your site, you hand it one markdown file at /llms.txt.
The format is deliberately small. An H1 naming the project or site is the only required section. Under it goes a blockquote with a short summary, then any number of H2 sections listing markdown links with a one line note each, plus a section called Optional for material that can be skipped when context runs short. The proposal also suggests clean markdown versions of important pages, at page.md or page.html.md.
Two consequences of that design get lost in most of the advice written about the file:
- It is read on demand, not crawled as a rule. The proposal's own framing is that a sitemap serves search engines while llms.txt offers a curated overview for LLMs. It brings nobody to your site: it pays off only when something already fetching your content decides to read it.
- It says nothing about access. robots.txt decides which bots may fetch which paths. llms.txt grants no permission and breaks no rule when it is ignored.
Does Google use llms.txt?
No, and Google has now said so twice in plain language. Its guide to optimizing for generative AI features on Google Search states that you do not need to create new machine readable files, AI text files, markup or markdown to appear in Google Search, including its generative AI capabilities, and that doing so will neither harm nor help your site's visibility or rankings. A Search Central changelog entry dated 15 June 2026 added that clarification for llms.txt specifically, while noting it is fine to keep the file for other services or systems that use it.
The same guide is just as direct about the other thing people bolt on for AI: structured data is not required for generative AI search and there is no special schema.org markup to add, although it remains worthwhile because it makes you eligible for rich results in Search. And on the bigger question, Google's answer is that SEO best practices continue to be relevant, because its generative AI features are rooted in its core Search ranking and quality systems.
So the answer to the question in the title is no, with no asterisk. If a checker marks your site down for a missing llms.txt and calls it a ranking factor, that claim has no primary source behind it.
Do ChatGPT, Claude and Perplexity read it?
Not as a documented input. Every major provider publishes crawler documentation, and what those pages describe is robots.txt: which user agent does what, and how to allow or block each one. OpenAI documents OAI-SearchBot for its search features, GPTBot for training foundation models, ChatGPT-User for actions a person starts inside ChatGPT (noting that because a user starts them, robots.txt rules may not apply) and OAI-AdsBot for ad safety checks. Anthropic documents ClaudeBot, Claude-SearchBot and Claude-User the same way, and states that its bots respect robots.txt. Perplexity documents PerplexityBot for its search results and notes that Perplexity-User, which a person triggers, generally ignores robots.txt. None of that documentation promises to fetch an llms.txt file or to treat it as a signal.
There is a pleasant irony here: OpenAI publishes an llms.txt of its own, and it shows the convention doing the job it was designed for. It indexes the developer documentation, with a blockquote summary and H2 sections pointing to per-product indexes, so an agent already working on the OpenAI API can pull the right page instead of walking the docs site. That is a convenience for tools that came looking, not a way to get found.
Then why publish one?
Two reasons survive scrutiny. The first is documentation an agent will actually fetch: API references, product docs, developer guides, where an assistant is asked to work with your material and a clean index saves it from guessing. The second is that writing the file is a useful exercise, because a one paragraph summary plus a short list of the pages that matter, one line each, is the same discipline that makes a site legible to people. The cost is close to zero, and Google has said keeping the file does no harm.
The argument against is maintenance: an llms.txt file is a second copy of your link list, and second copies drift. A file pointing at pages you renamed six months ago is worse than no file at all.
What actually decides whether AI answers mention you?
The boring work, in this order.
- Let the search crawlers fetch your pages. Each provider names its own search crawler (OAI-SearchBot, Claude-SearchBot, PerplexityBot) and documents robots.txt as the control. A page an assistant cannot fetch cannot be cited by it, whatever your llms.txt says.
- Serve real HTML. Google's advice for its AI features starts with a clear technical structure and unique, valuable content, and the most common way to fail that is a page whose text only exists after client side JavaScript has run. Prerendering or server side rendering of your public pages is worth more than any AI specific file.
- Answer first, in the page itself. Open each important page with a self contained answer of two or three sentences and use question shaped headings, so a model has something quotable to lift. See how AI models choose which brands to cite.
- Keep one description of your brand everywhere, and ship the ordinary basics: sitemap, a robots.txt that does not block the AI search crawlers, and structured data as part of your normal SEO work.
That list is GEO in practice, and it overlaps heavily with what SEO already asked of you. Once it is in place, the question becomes measurement rather than markup: how to check if ChatGPT mentions your brand covers the manual method, and Huntair SEO runs 200+ SEO and GEO checks on your site, with paid plans tracking whether ChatGPT, Gemini, Claude and Perplexity mention your brand for the prompts you choose.
If you publish one, keep it correct
A short checklist, straight from the format:
- Start with an H1 naming the site, then a blockquote summary. Everything else is optional.
- Group links under H2 headings, one line of context per link, and put anything skippable under Optional.
- Link only URLs that exist and return 200, and review the file whenever you rename or retire a page.
- Serve it as plain text at
/llms.txt, and never present it as a ranking or citation factor.
FAQ
Is llms.txt the same as robots.txt?
No. robots.txt is a widely supported access file: OpenAI, Anthropic and Perplexity all document it as the way to allow or block their crawlers, bot by bot. llms.txt is a content index with no access rules in it, and no provider is committed to reading it. If you only have time for one file, it is robots.txt.
Can an llms.txt file hurt my Google rankings?
No. Google states that creating machine readable files of this kind will neither harm nor help your visibility or rankings in Google Search, and its June 2026 changelog repeated that for llms.txt specifically. The only real risk is opportunity cost: time spent on the file instead of on crawler access, rendering and content.
Will llms.txt get my brand cited by ChatGPT?
There is no evidence that it will. Citation depends on a search crawler being allowed to fetch your page, the page being readable without JavaScript, and the content holding a clear answer worth quoting. An index file changes none of those three things.
Do I need Markdown versions of my pages too?
Only if you have reference material that a tool you care about will fetch. The proposal suggests a clean markdown copy of useful pages at page.md or page.html.md, which suits documentation. For a marketing site, server rendered HTML already gives every crawler what it needs.