llms.txt will not get you cited. Here is the data.
2 min read Augustyn Głowacki
llms.txt does not improve AI citations, and the evidence against it is now large enough to state flatly. Ahrefs checked server logs across 137,210 domains with May 2026 traffic: 28% of those domains publish an llms.txt, and 97% of those files received zero requests that month - no bots, no humans, nothing (Ahrefs, 15 June 2026). Separately, SE Ranking tested roughly 300,000 domains for a correlation between having an llms.txt and getting cited by an LLM, and found none - removing the variable from their prediction model actually improved its accuracy (SE Ranking, 7 November 2025). Google says the same thing in its own documentation, without qualification.
What does Google actually say about llms.txt?
“You don’t need to create new machine readable files, AI text files, markup, or Markdown to appear in Google Search (including its generative AI capabilities), as Google Search itself doesn’t use them” (Google AI optimization guide). That’s a direct statement from the one search engine that publishes detailed guidance on this question, and it isn’t hedged.
What did John Mueller say about it?
Reconciling that guidance against Chrome’s Lighthouse team shipping an experimental llms.txt check in its Agentic Browsing audits, Google’s John Mueller called llms.txt “not done for search” - a “temporary crutch, perhaps to save some tokens” for AI coding tools parsing developer documentation, not a search or citation mechanism (quoted in Ahrefs’ study and covered by Search Engine Journal). No engine - not OpenAI, not Anthropic, not Perplexity - has ever documented llms.txt as a retrieval or citation input.
Who is actually requesting these files, when anyone does?
Mostly not AI. Of the 3% of llms.txt files that did get fetched in Ahrefs’ study, 96% of requests came from bots, and 77% of those bots weren’t AI tools at all - Slackbot’s link-preview fetcher hit llms.txt more often than PerplexityBot did. Another 12% of fetches were tools studying llms.txt files, not consuming them for any downstream purpose. And zero AI bots probed for an llms.txt file that didn’t exist, which is the behavior you’d expect if any crawler treated the file’s presence as informative.
So what is llms.txt actually for?
Orienting coding agents and IDE tools around developer documentation - the use case Mueller named directly. Perplexity’s own documentation site serves an llms.txt for exactly that reason. If your content already lives as markdown - which it does, if you’re using the no-CMS approach I build - shipping a correctly-scoped llms.txt costs nothing extra and helps a coding agent working in your docs. It is not, and should not be sold as, a lever for getting cited by ChatGPT, Perplexity, or Google AI Overviews.
What should you check instead?
Whether the crawlers that actually decide citation can reach your pages at all, at robots.txt, at your CDN, and at your WAF. GPTBot trains models and doesn’t feed answers; the bots that decide what gets cited are the search-and-retrieval ones - OAI-SearchBot, Claude-SearchBot, Claude-User, PerplexityBot. Ahrefs frames the contrast in one line worth keeping: “llms.txt might not impact your visibility online, but robots.txt definitely does.” Check that file first, not llms.txt.