Does llms.txt work? What the evidence says

llms.txt is a text file that sits in a site's root directory and aims to present the site's most important pages to language models as a plain list; it was proposed in 2024 and is a proposal, not a standard. On today's evidence it is not a tool that increases visibility in AI answers: Google writes that Search does not use these files, and the largest published log study shows that the vast majority of the files are never requested. Publishing one is harmless, but your time is better spent on other work.

This article is not a measurement report. We did not measure the effect of llms.txt ourselves; every statement below rests on a published source, and the sources were re-read on 7 October 2026. We have tried not to say anything a source does not say, and to report what it does say together with its limits.

What is llms.txt?

llms.txt is a proposal Jeremy Howard published on the llmstxt.org site on 3 September 2024. The aim is for an AI agent to reach a site's concise, expert-level information in one place, instead of parsing HTML pages full of menus, ads and scripts.

The format of the proposal is simple. The file sits at the site root (/llms.txt) or under a subpath (such as /docs/llms.txt) and is written in a lightweight markup format that people can also read (Markdown): a heading, an optional short summary, then annotated lists of links grouped into sections.

For a fictional bicycle shop, an llms.txt might look like this:

# Örnek Bisiklet

> A shop in İzmir that sells city and mountain bikes and offers servicing and spare parts. This file lists the most important pages on the site.

## Products

- [City bikes](https://www.ornek.com.tr/sehir-bisikletleri): models, frame sizes, price list
- [Mountain bikes](https://www.ornek.com.tr/dag-bisikletleri): models and technical specifications

## Services

- [Servicing and maintenance](https://www.ornek.com.tr/servis): maintenance packages, appointments
- [Returns and warranty](https://www.ornek.com.tr/iade-garanti): terms and time limits

## Company

- [About us](https://www.ornek.com.tr/hakkimizda): year founded, address, contact

It should not be confused with robots.txt. robots.txt tells bots what you ask them not to crawl, and companies explain in their own documentation whether they comply; details are in our robots.txt guide. llms.txt, by contrast, says what you suggest they read, and no AI search bot is documented as committing to read it.

What does Google say?

Google has taken a clear, written position on this. According to the guide Optimizing your website for generative AI features on Google Search, to appear in Google Search, including its AI features, you do not need to create new machine-readable files, AI text files or special markup, because Google Search does not use them.

The same guide goes one step further: it writes that publishing an llms.txt file for other services that use such files is perfectly fine, but that this will neither harm nor help visibility or ranking in Google Search, because Google Search ignores these files. According to Google's documentation updates page, this note was added on 15 June 2026. The guide also lists creating llms.txt among the "AEO/GEO hacks" you can ignore for Google Search.

The older AI features and your website document points the same way: it says there are no additional requirements or special optimizations for appearing in Google AI Overviews or AI Mode, and that you do not need to create new machine-readable files or AI text files.

A statement in the same direction came from Google even before the official documents. As reported by Search Engine Roundtable on 18 June 2025, Google's John Mueller wrote on Bluesky that no AI system currently uses llms.txt, adding that a look at server logs shows bots do not request the file. That is one person's observation at that date; for Google's current position, the guide above is the reference.

One limit is worth noting: Google's statement applies only to Google Search and its AI features. It does not speak for ChatGPT, Claude or Perplexity. For those engines, the best information we have is the log data below.

Do bots actually request the file?

The first step in finding out whether a file works is to check whether anyone requests it. The largest study on this that explains its method is the analysis Ahrefs published on 15 June 2026.

According to Ahrefs, the search bots of OpenAI, Perplexity and Claude sent only a few hundred requests to these files in total. The study concludes that if the goal is to appear in ChatGPT, Perplexity or AI Overviews, llms.txt is largely decorative.

The article also states the study's limits. Ahrefs Web Analytics customers are more technical and more SEO-minded than the average website, so the 28% publishing rate should be read as an upper bound. The measurement covers a single month and counts requests, not citations in AI answers.

Next step

Is your site open to AI crawlers?

Enter your domain. The free scan measures whether AI crawlers can reach your site and whether your pages are quotable. No signup, no card, no email.

Your domainScan for free

What does the evidence not show?

The strength of the evidence above needs to be separated from its limits, because "no evidence" and "proven not to work" are not the same thing.

• What it shows: Google Search does not use the file; Google says so itself. In the largest published log study, the vast majority of files were never requested over a month, and AI search bots sent very few requests.
• What it does not show: That the file will never be read by any system. AI coding agents, or teams building their own tools, may find a documentation index useful; offering AI agents the essence of a site in one place was the aim of the proposal in the first place.
• What we do not know: Whether llms.txt changes how often a brand is mentioned or cited in ChatGPT, Gemini, Perplexity, Claude or AI Overviews answers. We found no study that measures this and publishes its method. We have not measured it either.

So the conclusion of this article is not a verdict but a reading of today's evidence. If an engine documents that it reads llms.txt, or a measurement with a transparent method shows an effect, we will update this article.

What should you do?

Publishing llms.txt is harmless; even Google writes that it will not harm Search. If you publish one, watch two things: the file must really be plain text, and the information in it must not contradict the pages on your site.

The first point comes up more often than you might think. On sites built as single-page applications, every non-existent address may return the site's home page with a 200 code. In that case the /llms.txt address appears to "exist", but what it holds is an HTML page. Open the address in your browser: if you see your site's design, the file does not exist.

On the current evidence, the work that determines AI visibility lies elsewhere:

1. Bot access. If search bots cannot reach your page, nothing else matters. robots.txt rules and CDN settings are the first places to check.
2. Server-rendered content. According to network data Vercel published in December 2024, the OpenAI, Anthropic and Perplexity bots examined did not execute JavaScript, while Gemini could process JavaScript through Googlebot's infrastructure. A page whose content is built only in the browser with JavaScript may look empty to these bots.
3. Citable pages. Pages that give a clear, short, sourced answer to the question asked, and carry basic information such as product details, prices and contact details as text. You can find what to check, item by item, in our checklist.

Our scan, which requires no sign-up, checks these three areas and also reads llms.txt, but as an optional item: a missing file does not lower the score. The main finding the scan shows for llms.txt is when an HTML page is served at that address instead of the file. And if we cannot read the file, we do not say "no file"; we mark the item as not measurable.

Frequently asked questions

Will publishing llms.txt make me appear in ChatGPT or AI Overviews?

On today's evidence, no. Google writes that Search and its AI features do not use llms.txt. For ChatGPT and the other engines, we found no study with a transparent method showing that the file changes visibility in answers.

Will publishing llms.txt harm my site?

According to Google's guide, no: the file neither harms nor helps visibility or ranking in Google Search. The only risks to watch are the information in the file contradicting the pages on your site, or an HTML page being served at the address instead of the file.

What is the difference between llms.txt and robots.txt?

robots.txt tells bots what you ask them not to crawl; OpenAI, Anthropic, Perplexity and Google recommend in their own documentation that you manage their search and training bots with this file. llms.txt is a list that says what you suggest they read, and there is no documented commitment that it will be read. For visibility, getting robots.txt right matters far more.

Do you have your own measurement on this?

No. We have not measured the effect of llms.txt on AI answers; this article brings together published documents and log data. What we do measure is whether the file is really served as plain text on your site.