Does llms.txt get your site cited by AI?
No. No published study has shown that adding llms.txt changes whether AI engines cite you, and the engines that have commented on it say they do not use it for ranking or citation. It takes about an hour to add and does no harm. It is not a citation strategy.
Verdict: no evidence of benefit Confidence: moderate–high
The claim
llms.txt is a plain text file you put at the root of your site. It was proposed in 2024 as a kind of robots.txt for AI: a map that tells language models which pages matter and how to read them. The pitch was intuitive, the file is trivial to write, and by mid-2025 tooling had made it a one-click job.
The claim attached to it, repeated across hundreds of articles, is that adding the file improves your odds of being cited by ChatGPT, Perplexity, and Google’s AI surfaces.
The receipts
We have been collecting everything the GEO industry publishes, every day. Here is what the record actually shows about this claim, in order.
- Mar 2025 “The new SEO hack for AI-powered search.” quoleady.com — the file enters the mainstream playbook.
- Apr 2025 “What is llms.txt, and should you care about it?” ahrefs.com — a major SEO brand starts hedging early.
- Jun 2025 Yoast ships llms.txt as a product feature. yoast.com — adoption stops being a decision and becomes a default.
- Sep 2025 “Largely ignored: why it has no impact on AI search or GEO.” promptwatch.com — the first published doubt.
- Sep 2025 “The simple file that could double your AI citations.” maximuslabs.ai — published the same month as the finding above.
- Nov 2025 “Why brands rely on it and why it doesn’t work.” seranking.com
- 21 Nov 2025 “New study finds llms.txt has no clear impact on AI citations.” stanventures.com — the point where the evidence became hard to miss.
- Feb 2026 “Why every website needs one in 2026.” aether-agency.co.uk — five months after the first doubt.
- Feb 2026 “844K+ sites already have it.” loamly.ai — adoption cited as its own justification.
- Mar 2026 “Why every website needs it in 2026.” nicodigital.com — six months after the first doubt.
of the 39 articles in our corpus that actively promote llms.txt were published after the first article reporting that it does nothing. They came from 28 different publishers.
That gap is the actual story. Doubt was published on 12 September 2025. Promotion carried on for another nine months from 28 separate publishers, the most recent we hold dated June 2026. Put plainly: almost nine in ten of the promotional pieces we have were written after somebody had already shown the file does nothing.
It is the clearest example we have of a habit worth naming. This industry does not retract. A claim gets published, the finding that undercuts it gets published, and the claim keeps circulating anyway — because nobody is keeping the record.
How we classified. Every title in our llms.txt set was sorted into promotes / doubts / neutral, and we counted how many promotional pieces came after the first published doubt. Titles can mislead, so treat the exact figure as indicative rather than surgical. The direction is not in question: the promotional stream never slowed.
What this means for you
- Already have one? Leave it. It is harmless, it costs nothing to keep, and removing it buys you nothing. This is not a cleanup task.
- Don’t spend a sprint on it. If a vendor’s AI-visibility plan leads with llms.txt, that tells you something useful about how current their thinking is.
- Spend the hour on being mentioned instead. The consistent finding across citation studies is that being referenced on sources the engines already trust moves the needle far more than anything you can configure on your own domain.
Confidence, and what would change our mind
Moderate to high that there is no citation benefit today. Multiple independent analyses point the same way, and no engine has claimed otherwise.
Low that this stays true forever. The file is cheap and the agent tooling landscape is moving fast. We would revise this page if:
- OpenAI, Google, Perplexity, or Anthropic states that llms.txt is used in retrieval or citation selection.
- A controlled study with a proper holdout group shows a citation lift.
- Agent frameworks begin fetching it at inference time in measurable volume — which would make it useful for a completely different reason than the one it was sold on.
How we know
We run a crawler that collects GEO and AI-search publishing continuously and stores it with dates. As of this review it holds 19,293 articles, of which 122 across 77 publishers name llms.txt in the headline — earliest dated June 2024, most recent July 2026.
The timeline above is drawn from those records, not from memory. Every source is linked with its date so you can check our reading against it. When we revise this page we date the revision rather than quietly editing it, which is the whole point of keeping a record in the first place.
Sources
- 31 Mar 2025 · quoleady.com LLMs.txt: The New SEO Hack for AI-Powered Search
- 22 Apr 2025 · ahrefs.com What Is llms.txt, and Should You Care About It?
- 10 Jun 2025 · yoast.com llms.txt — Yoast SEO Features
- 12 Sep 2025 · promptwatch.com LLMs.txt Is Largely Ignored: Why It Has No Impact on AI Search or GEO
- 20 Sep 2025 · maximuslabs.ai llms.txt Implementation: The Simple File That Could Double Your AI Citations
- 9 Oct 2025 · presenceai.app LLMs.txt Reality Check: AI Search Impact & Agent Use Cases
- 7 Nov 2025 · seranking.com LLMs.txt: Why Brands Rely On It and Why It Doesn’t Work
- 21 Nov 2025 · stanventures.com New Study Finds LLMs.txt Has No Clear Impact on AI Citations
- 11 Feb 2026 · aether-agency.co.uk What Is llms.txt and Why Every Website Needs One in 2026
- 18 Feb 2026 · loamly.ai llms.txt: What It Is, How to Create One, and Why 844K+ Sites Already Have It
- 8 Mar 2026 · nicodigital.com What Is LLMs.txt? Why Every Website Needs It in 2026