# What is llms.txt? > llms.txt is a Markdown file, usually at your domain root, telling AI models which pages matter: what it holds, who reads it, what V2 changed, and its limits. Source: https://llmstextgenerator.com/what-is-llms-txt/ Last updated: 23 October 2026 A Markdown file, usually at the root of your domain, that tells language models what your site publishes and which pages matter. It takes about twenty minutes to write and roughly a minute to generate. **llms.txt is a plain-text index for AI readers.** It sits at `https://yourdomain.com/llms.txt`, names your site, summarises what you do in a sentence or two, and lists your most useful pages with a one-line description of each. It is written in Markdown, so people can read it as easily as machines. ## What is llms.txt used for? When a language model needs to know something about your business, it does what you would do: it fetches a page and reads it. That page arrives as HTML full of navigation, cookie banners, scripts, footers and markup. Somewhere inside is the paragraph that actually answers the question. Models are reasonably good at finding it, but two things work against them. Context windows are finite, and boilerplate uses up space that could have held your content. More importantly, nothing in the HTML says which of your two hundred pages is the one worth reading. A model landing on your careers page has no signal that your documentation exists. An `llms.txt` file solves that second problem. It is a short, curated index in plain Markdown, written for machines but perfectly readable by people. ## What does an llms.txt file look like? There is no new syntax. An H1 for your site name, a blockquote summary, then H2 sections containing lists of links, each with a one-line description. llms.txt ``` # Northgate Tools > Independent supplier of woodworking hand tools, based in Leeds. We sell, > sharpen and restore. Orders ship within two working days. ## Products - [Chisels](https://example.com/chisels): Bevel-edge and mortise chisels from six makers, 6mm to 38mm. - [Hand planes](https://example.com/planes): New and restored bench planes, block planes and spokeshaves. ## Services - [Sharpening](https://example.com/sharpening): Postal sharpening service, GBP 6 per edge, 3-day turnaround. - [Restoration](https://example.com/restoration): Full restoration of vintage tools. Quote required. ## Optional - [Our workshop](https://example.com/about): History of the shop and the people in it. ``` Publish that at `https://yourdomain.com/llms.txt` and you are done. The [format guide](https://llmstextgenerator.com/llms-txt-format/) covers each element in detail. ## Where did the idea come from? Jeremy Howard of Answer.AI proposed the file in September 2024. The reasoning was simple: the web is built for browsers, reading it is expensive for a model, and a small curated file is a cheap way to hand over the good parts directly. You can read [the llms.txt proposal at llmstxt.org](https://llmstxt.org/) and follow its development in [the proposal’s GitHub repository](https://github.com/AnswerDotAI/llms-txt). In August 2026 the proposal was revised as V2. It kept the format and added standard ways for a page to point at its llms.txt and at a Markdown version of itself. The [V2 guide](https://llmstextgenerator.com/llms-txt-v2/) covers what changed. **It is a proposal, not a standard.** No standards body has ratified `llms.txt`, and no AI company is obliged to read it. Adoption is real, especially among coding tools and documentation platforms, but uneven. Anyone telling you it is required is selling something. ## What does llms.txt not do? - **It does not control access.** That is [`robots.txt`](https://llmstextgenerator.com/llms-txt-vs-robots-txt/). Listing a page in `llms.txt` grants no permission it did not already have, and leaving one out blocks nothing. - **It does not affect search rankings.** Google says you do not need to create AI text files to appear in its AI features ([Google’s guidance on AI features](https://developers.google.com/search/docs/appearance/ai-features)). [The longer answer](https://llmstextgenerator.com/llms-txt-for-seo/) is worth reading before you make claims to a client. - **It does not replace your sitemap.** Search engines still want `sitemap.xml`. The two coexist. - **It is not AGENTS.md.** That file guides coding agents inside a code repository; [llms.txt vs AGENTS.md](https://llmstextgenerator.com/llms-txt-vs-agents-md/) explains the difference. - **It does not guarantee anyone reads it.** You are publishing an offer, not issuing an instruction. ## Is it worth doing? For most sites, yes, on cost rather than hype. It is one small file, it takes an hour at most, it cannot hurt you, and when a model does describe your business it can do so from a summary you wrote rather than from whatever it scraped off your homepage. The V2 notes put it plainly: coding agents now use these files reliably. That is a reasonable return even if general assistants are slower to follow. The honest caveat is that the benefit is not measurable today. You will not see it in analytics. If you need proven returns before acting, this is not that; if you are comfortable with cheap insurance, it is. [Whether your site needs one](https://llmstextgenerator.com/do-i-need-llms-txt/) goes through the cases where it clearly helps and where it is a waste of time. ## Frequently asked questions Is llms.txt an official standard? No. It is a public proposal, first published in 2024 and revised as V2 in August 2026, and adopted voluntarily. There is no governing body and no compliance requirement. Do I need one if I already have a sitemap? They do different jobs. A sitemap lists every URL for crawlers; llms.txt picks out the important ones and says what each is about. Keep both. Can it hurt my site? The only realistic risk is publishing something inaccurate, or listing a page you did not mean to draw attention to. Review the file before it goes live and you are fine. What size should it be? Small. Most good files are 2 to 10 KB. If yours is approaching a hundred kilobytes you have written a sitemap, not an index.