By Priya Raman, AI SEO, GEO & Martech Analytics. Last updated: 2026-10-23
What is llms.txt, in plain terms?
The proposal asks sites to publish a Markdown file at /llms.txt containing the site's name, a short summary, and curated links to the pages a model should read, with an expanded variant often called llms-full.txt. The analogy is robots.txt, and the analogy is also the trap: robots.txt works because crawlers agreed to honour it. A file only matters if the systems it addresses actually fetch and use it, which is exactly what the studies below tested.
What does the evidence show?
Three independent checks point the same way, and this post reports them as they were published, with their samples stated.
- Google's stated position. Google says Google Search does not use llms.txt, and that no special files are needed for AI Overviews or AI Mode. Treat any claim that the file is required for Google AI surfaces as contradicted by the platform itself.
- Ahrefs log study, May 2026. Ahrefs examined about 137,000 domains in its Web Analytics set, found about 28 percent had published an llms.txt file, and found 97 percent of those files received zero requests during May 2026. Of the requests that did occur, most came from bots generally, with AI retrieval bots a small share.
- SE Ranking correlation analysis. SE Ranking tested whether having an llms.txt file correlated with how often a domain is cited by AI systems across about 300,000 domains, and found no correlation.
None of this proves the file causes harm, and none of the sources claim it does. The finding is narrower and, for a budget holder, decisive: no published mechanism or dataset currently shows the file increasing AI visibility or citations.
Want to put this into practice?
Reachium automates LinkedIn outreach, content publishing, and inbox management in one platform.
See Reachium →So should a B2B marketing team publish one?
If your team can generate an accurate file cheaply and keep it accurate, it is harmless housekeeping, and it may genuinely help where the audience is technical: coding agents and developer tools are documented consumers of site documentation files, and vendors such as documentation platforms can generate the file automatically. Those are workflow reasons.
What the file should never be is a line item sold as AI search optimization, a substitute for crawlable pages, or a reason to delay the work covered in how to measure AI search visibility and how to get cited by ChatGPT: pages that answer buyer questions directly, structured so an engine can quote them, and present in the search results engines retrieve from. Our GEO content structure guide covers that page-level work.
FAQ
Does llms.txt help AI search visibility? No published evidence says it does. Google says Search ignores it, the Ahrefs log data shows the files are rarely fetched, and SE Ranking found no citation correlation. That is absence of evidence at unusually large sample sizes, which is as close to a verdict as this category currently offers.
Do any AI systems actually read llms.txt? Coding agents and some developer tools do use site documentation files, and AI crawlers occasionally fetch the file, per the Ahrefs logs. The answer engines' retrieval bots, the ones that would need to read it for visibility to change, were a small share of observed requests.
What should a good llms.txt contain, if we publish one? Follow the proposal at llmstxt.org: the site name as the heading, a short factual summary, and a curated list of your most important pages with brief descriptions. Keep it accurate. A stale or inflated file is worse than none if an agent does read it.
Is llms.txt the same as the keywords meta tag? No, but the comparison is made often because both are self-descriptions an engine is free to ignore. The useful lesson is the same: engines trust what they can verify on your pages, not what you declare about yourself in a side file.
Will publishing one hurt our rankings? Google's stated position is that the file neither helps nor harms Search visibility, since Search ignores it. The real cost is opportunity cost: hours spent polishing a file nobody fetches are hours not spent on pages that earn citations.
Sources
- The llms.txt proposal, llmstxt.org, Jeremy Howard, first published 2024-09-03.
- Google Search guidance on llms.txt as reported in 2026 coverage: Google Search ignores the file, with no effect on rankings, AI Overviews, or AI Mode.
- Ahrefs, llms.txt access study, published 2026-06-15: about 137,210 domains, May 2026 logs, 97 percent of published files received zero requests.
- SE Ranking, llms.txt and AI citation analysis: about 300,000 domains, no correlation found between having the file and AI citation frequency.
- GTMStack queue of record, .saeo/keywords.yaml, entry "llms.txt", research sampled 2026-10-21.
- Internal link targets: /get-cited-by-chatgpt and /geo-content-structure (live); /how-to-measure-ai-search-visibility (staged in the same batch).
Linked sources:
