llms.txt is a plain Markdown file, proposed in 2024, that a website publishes at its root to hand AI systems a short, curated map of its most important pages. The honest answer to whether it helps you get found by AI: not right now, and not by the engines that decide AI visibility. Google says directly it doesn't use it. Server-log studies show the crawlers behind ChatGPT and Claude barely request it. It isn't harmful to publish — it just isn't a GEO lever. The signals that actually decide whether AI can find and cite a business are covered on the what is GEO pillar, not in this file.
llms.txt is a Markdown file, published at a site's root, with an H1 project name, a one-line summary, and H2 sections linking to a site's most important pages. Jeremy Howard and the Answer.AI team proposed it in September 2024 as a fix for a real problem: most websites are too large and too full of navigation, ads and JavaScript for a language model to read efficiently, so a short, hand-curated index was meant to give models a faster path to the content that matters. A companion file, llms-full.txt, holds the full text inline for models that can take it.
Proof: the specification and its background are published by its author at llmstxt.org.
Bottom line: it's a real, well-specified format. The open question was never whether it exists — it's whether any engine that matters actually reads it.
Google has stated plainly that it does not support llms.txt and has no plans to. Google's Gary Illyes confirmed this at a Search Central Live event, and Google's own AI-optimization guidance states that no special AI text file is needed to appear in generative AI search features — ordinary indexability and quality signals are what those features build on, the same ones this site's GEO pillar walks through.
Do this: treat any claim that llms.txt boosts AI Overviews or Google's AI features as false until Google says otherwise — right now it says the opposite.
The retrieval crawlers that could act on llms.txt mostly don't request it — that's now measured, not guessed. Ahrefs analyzed server logs across roughly 137,000 domains and found that 97% of published llms.txt files received zero requests in the period studied. Of the small share that were fetched, most hits came from SEO audit tools checking whether a file existed — not from the crawlers that actually retrieve content for AI answers. GPTBot accounted for a small single-digit share of llms.txt requests and ClaudeBot for under 1%.
Proof: Ahrefs — server-log study of llms.txt requests across 137,000 domains.
Bottom line: adoption isn't the problem — reading it is. A file almost nobody's crawler requests can't be influencing what gets cited.
llms.txt looks like robots.txt and sitemap.xml, two files that genuinely do matter to AI crawlers — but looking similar isn't the same as being adopted the same way. Robots.txt controls whether GPTBot, ClaudeBot, Google-Extended and PerplexityBot can reach a site at all; blocking one by accident, often through a CDN's default bot rule, is one of the most common AI-visibility failures. Sitemap.xml genuinely aids discovery. llms.txt has the same file-at-the-root shape, but no engine's own documentation instructs its crawlers to fetch or act on it as part of retrieval or citation.
Rule of thumb: a file's format looking like a standard doesn't make it one — check whether the engine's own documentation names it, the way it does for robots.txt.
Publishing llms.txt won't hurt, and it has a couple of narrow, honest uses — just not AI-search citation. Anthropic and several developer-tool companies publish their own llms.txt files on their documentation sites, mainly to help coding assistants and agent tools that are explicitly built to look for the file navigate dense docs — a different job from general AI-search visibility. It costs a few minutes to write and there's no real downside to having one.
Shortcut: if a site is documentation-heavy and talks to coding agents, five minutes on an llms.txt file is fine. For everyone chasing AI recommendations, spend that hour elsewhere.
The signals that decide AI visibility have nothing to do with a Markdown index file. As the what is GEO pillar covers in full, it comes down to crawler access, entity clarity, buying-intent content, first-hand evidence and independent corroboration — five things an engine actually checks before it will name a business in an answer.
Do this: before writing an llms.txt file, confirm none of the AI crawlers are blocked in robots.txt — that single check affects visibility far more than any curated index ever will.
The free check reads your crawler access and entity clarity — 3 minutes, evidence included.