llms.txt is a proposed plain-text file, written in Markdown, that lists your most important pages so AI tools can find them. It is not an official standard, Google does not use it for search, and in mid-2026 there is no measured evidence it improves your odds of being cited by ChatGPT, Claude, Gemini, or Perplexity. It does have a real use for one group: teams maintaining developer documentation that AI coding tools read. For most service businesses, it is optional and low-priority.
The honest version of this answer matters because the marketing around llms.txt has outrun the evidence. Vendors sell it as an AI-visibility unlock. The data and the platform owners say otherwise. Here is what is actually true.
What does an llms.txt file actually contain?
An llms.txt file is a Markdown document at your domain root that uses H2 headers to group links to your key pages. The format is simple: a title, an optional summary of what the site is, then sections of curated links with short descriptions. The idea, proposed by Jeremy Howard in 2024, is to hand a language model a clean, token-efficient map of your site instead of making it crawl and parse your full HTML. The motivation was practical: HTML pages carry navigation, scripts, and styling that waste tokens and obscure the actual content, so a stripped-down Markdown index would, in theory, let a model find what matters faster and cheaper.
There are two flavors in practice. A basic llms.txt points to the pages you want models to prioritize. An expanded approach, sometimes published as llms-full.txt, dumps clean Markdown versions of the content itself so a model can read the text without rendering a page. Ahrefs describes the file as a curated index, not a permissions file, which is the key distinction from robots.txt: llms.txt suggests what to read, it does not control access.
Does Google use llms.txt?
No. Google’s Search team has said it does not use or endorse llms.txt. In July 2025, Google’s Gary Illyes confirmed Google does not support the file and has no plans to, and John Mueller publicly compared it to the keywords meta tag, the discredited 1990s tag that search engines stopped trusting because it was trivial to abuse. Mueller’s point was blunt: llms.txt cannot help an LLM differentiate your site, and it is “not done for search.”
That comparison is the tell. The keywords meta tag failed because it was self-reported and unverifiable, so engines ignored it in favor of signals they could confirm independently. llms.txt has the same structural weakness. A file where you declare your own most important pages is easy to game and hard to trust, so the engines that decide citations lean on signals they can corroborate across the open web instead. If you want to understand what those signals are, start with what sources AI engines cite.
Do AI engines use llms.txt to decide citations?
There is no evidence the major answer engines use llms.txt to decide who gets cited. As of mid-2026, having the file does not measurably improve your odds of appearing in ChatGPT, Claude, Gemini, or Perplexity answers. The companies behind those products have not confirmed reading it for their search surfaces, and independent tests have not shown a citation lift tied to its presence.
This is the claim to be skeptical of when an agency pitches llms.txt as an AEO must-have. The things that do move AI citations are well documented: direct-answer content structure, statistics and original data, entity consistency across the web, and earned mentions in trusted sources. A self-published index file is none of those. If your AEO budget is finite, llms.txt sits far below the work that actually gets pages cited, which we lay out in what actually gets cited in AI search.
How many sites have adopted llms.txt?
Adoption is low and not climbing fast. An SE Ranking study of 300,000 domains found a 10.13% adoption rate, so roughly one site in ten carries the file after about eighteen months of industry discussion. That is meaningful awareness but not the kind of universal adoption you would expect if the file delivered a clear ranking or citation benefit.
The flat adoption curve is itself a signal. Standards that work tend to spread because early adopters see results and others copy them. llms.txt has had time and attention without a corresponding surge, which lines up with the lack of measured benefit. It is a proposal that the market is testing and mostly finding inessential.
When is llms.txt actually worth shipping?
It is worth shipping if you maintain developer documentation that AI coding tools consume. IDE agents like Cursor, Claude Code, GitHub Copilot, and Windsurf, along with some MCP servers and in-product AI assistants, fetch llms.txt to pull clean docs efficiently. If your users build against your API or library and ask their coding assistant about it, a well-formed llms.txt can make those tools answer more accurately about your product. That is a real, narrow benefit.
For everyone else, the calculus is different. The file is cheap to create, so shipping one does no harm, and it can serve as a tidy internal index of your priority pages. Just do not expect it to move AI search visibility, and do not let it displace the work that does. Build it in an afternoon if you like the housekeeping, then spend your real effort on content structure, entity signals, and earned coverage. Our AI search optimization guide puts the high-impact work in order.
How does llms.txt compare to robots.txt and sitemaps?
llms.txt sits between robots.txt and a sitemap in function but carries less weight than either. robots.txt is an enforced access rule that crawlers, including the AI crawlers from OpenAI, Anthropic, and Perplexity, generally respect, so it actually controls who reads your site. An XML sitemap is a machine-readable list of your URLs that search engines use to discover pages, and Google has supported it for nearly two decades. llms.txt is newer, unenforced, and unsupported by the major search engines, which makes it the weakest of the three signals.
The practical reason this ordering matters is budget. Time spent perfecting an llms.txt file is time not spent on your robots.txt directives, which decide whether AI crawlers can reach your content at all, or on your sitemap, which still drives discovery. If your AI crawlers are blocked in robots.txt, no llms.txt will save you, and a missing or stale sitemap costs you discovery that the index file cannot replace. Get the two enforced, supported signals right first, then treat llms.txt as optional housekeeping rather than a priority. If your pages are not showing up in AI answers at all, the cause is far more likely a crawl or index problem than a missing index file, which we diagnose in why your website is not showing up in AI search.
Frequently asked questions
Is llms.txt the same as robots.txt? No. robots.txt controls which crawlers may access your pages and is widely respected. llms.txt is a suggested index of your important content for language models and controls nothing. One is an access rule, the other is a curated recommendation.
Will adding llms.txt get my site cited by ChatGPT? There is no evidence it will. As of mid-2026, the file does not measurably improve citation odds in ChatGPT, Claude, Gemini, or Perplexity. Content structure, original data, and entity consistency are what move citations.
Did Google say to use llms.txt? The opposite. Google confirmed it does not support llms.txt, and John Mueller compared it to the discredited keywords meta tag. Google does not use it for search.
Who should actually create an llms.txt file? Teams with developer documentation that AI coding tools read. IDE agents like Cursor and Claude Code fetch it to answer questions about your product accurately. That is the clearest real use case.
Does llms.txt hurt anything if I add it? No. It is a static text file with no downside beyond the time to maintain it. The risk is opportunity cost: treating it as an AEO priority when higher-impact work is going undone.
What is the difference between llms.txt and llms-full.txt? llms.txt is a curated index of links to your key pages, while llms-full.txt embeds clean Markdown versions of the content itself so a model can read the text without rendering a page. Both are part of the same proposal, and neither is supported by Google for search or shown to move AI citation odds in mid-2026.
If llms.txt does not help, what should I do instead? Spend the effort on signals engines actually use. Confirm AI crawlers can reach your pages in robots.txt, keep a current sitemap, structure each page to answer a specific question in its first lines, add accurate FAQ and Article schema, back claims with data, and earn mentions on trusted third-party sources. Those moves are documented to influence citations in a way llms.txt is not.
If you want a clear read on what is actually moving your AI visibility, skip the file and start with a free AI visibility analysis or contact us for an honest audit of where your citations come from.
Tagged