llms.txt is one of the most talked-about and most overhyped topics of the last two years. Promises like “add a file to your site and appear in AI” are circulating. Let me say this upfront: this is not a guarantee. Let's look separately at what it is, who uses it, and who does not say they use it.
What is llms.txt?
llms.txt is a file format proposed by Jeremy Howard on 3 September 2024. It sits in the site's root directory (/llms.txt) and lists the site's important pages in plain Markdown for language models and agents. The proposal's own page makes clear that it is not an internet standard but a proposal.
The format is simple: an H1 with the site name, a blockquote with a short summary, then lists of links with descriptions under headings; secondary links go under the “Optional” heading.
Who uses it?
- Google: In the official guidance it published on 15 May 2026, it states that Google Search does not use the llms.txt file, ignores it, and that creating one will neither help nor harm Google Search visibility.
- Anthropic, OpenAI, Perplexity: all three publish llms.txt on their own developer documentation sites. I could not verify an official statement that their products use this file in search or citation ranking.
- Microsoft: I could not verify an official statement that Bing/Copilot uses llms.txt.
So what do we do about it?
I see llms.txt as a low-cost preparation step: a single Markdown file, quick to prepare, and it can serve human readers and agents like a site map. But I never present it as a solution on its own; it is one part of the technical foundation (robots.txt, sitemap, structured data, crawlable and unique content). This site has an llms.txt file too.
How is it written?
# Brand Name > 2-3 sentences describing what the brand does, who it serves and where it is located. ## Services - [Service A](https://ornek.com/hizmet-a): one-sentence description. - [Service B](https://ornek.com/hizmet-b): one-sentence description. ## Company - [About](https://ornek.com/hakkimizda): who we are. - [Contact](https://ornek.com/iletisim): how to reach us. ## Optional - [Privacy Policy](https://ornek.com/gizlilik)Common mistakes: dumping the whole site (selecting for the agent has value too), not keeping the file up to date, linking to hidden or unpublished pages, using exaggerated marketing language.
What should and should not be expected?
- What can be expected: presenting your brand's important pages in one place, clearly; a readable summary for agents and developers.
- What should not be expected: a causal claim such as “llms.txt was added, AI visibility increased.” Ask anyone who claims this for evidence; I could not verify this effect and I do not promise it.
Frequently asked questions
Can adding llms.txt do harm?
According to Google, it neither harms nor helps Google Search visibility. As long as it contains no incorrect information and is kept up to date, the risk of harm is low.
Which pages should I list?
The core pages that describe the brand accurately: services, about, contact, the most valuable guides. Not all of them.
Sources
- The llms.txt proposal (Jeremy Howard)
- Google: Optimizing your website for generative AI features on Google Search
- Anthropic developer documentation llms.txt
- OpenAI developer documentation llms.txt
- Perplexity developer documentation llms.txt
Sources checked on 1 October 2026; official pages may change over time.
