How to Write an llms.txt File (And Whether You Actually Need One)

An llms.txt file is a plain text file at your site’s root that lists your most important pages for AI models to reference. It’s genuinely easy to add. Here’s the honest part: no major AI platform, not OpenAI, Google, Anthropic, or Meta, has confirmed it actually uses these files when crawling for search, and the largest crawl log study published so far found that 97% of published llms.txt files never get requested by an AI crawler at all. So here’s exactly how to write one, who is actually using it, and what the real evidence says about whether it is worth your time in 2026.

Sources: PPC Land, Rankability

What Is an llms.txt File?

It’s a plain text file, formatted in markdown, placed at your site’s root directory (yoursite.com/llms.txt). It lists your highest value pages with a one line description of what each one covers, organized under simple headings. The idea is that an AI crawler can read this one flattened file instead of parsing your full site (navigation, ads, JavaScript, and all) to figure out what you actually offer.

The proposal was published by Jeremy Howard, co founder of Answer.AI and fast.ai and a former Kaggle president, on September 3, 2024. The specification is genuinely simple: one H1 heading is the only required element. Everything else, a one line blockquote summary, free form detail sections, and H2 delimited lists of links with short descriptions, is optional. The official spec lives at llmstxt.org, and it defines two related files: llms.txt, a streamlined navigation view, and llms-full.txt, a single file containing your full documentation content in one place.

It is still a proposed standard, not an official one. No major LLM provider has confirmed their crawlers read or prioritize it. It is one small piece of the broader GEO picture. See our complete guide to generative engine optimization for the three pillars that actually move the needle, technical signals like this one included.

llms.txt by the Numbers

Adoption is growing fast off a small base. Total llms.txt instances (including the related llms-full.txt and ai.txt formats) reached an estimated 38,980 sites by May 2026, with the core llms.txt count alone at 36,120, an 8.8x increase from roughly 4,000 a year earlier. Among the web’s top 1,000 sites, adoption sits around 8.7% as of June 2026. In the top 10,000, adoption grew from roughly 1.04% in July 2025 to about 5.61% by June 2026, over a 5x increase in less than a year.

Adoption is not evenly distributed. Mid traffic sites (1,001 to 5,000 monthly visits) adopt at a higher rate, around 10.54%, than high traffic sites over 100,000 visits, which sit closer to 8.27%. By mid 2025, adoption was already routine among developer facing SaaS companies. Through the first quarter of 2026, it expanded into mainstream SaaS, publishing, and some consumer sectors, while financial services, healthcare, and legal, sectors with heavier compliance concerns, remained slow to adopt. Enterprise adoption specifically lags well behind the open web: only an estimated 7.4% of Fortune 500 companies, 37 out of 500, had implemented llms.txt as of March 2026.

llms.txt vs robots.txt vs sitemap.xml

These three files get confused constantly, but they do genuinely different jobs, and only one of them is actually enforced.

Robots.txt controls access using Allow and Disallow directives, and it cannot be ignored by a compliant crawler. llms.txt does not control access at all; you cannot block anything with it, and there is no evidence any major crawler is required to honor it, or does. Sitemap.xml exists to help search engines discover and index your pages, a job that predates AI search entirely. llms.txt is the only one of the three built specifically for language models, and it is also the only one with no confirmed enforcement behind it.

llms.txt vs llms-full.txt: What’s the Difference?

The spec defines two separate files, and the similar names cause constant confusion. They do different jobs:

• llms.txt is a curated index: a short markdown list of your most important pages with one-line descriptions. Think of it as a table of contents an AI model can skim.
• llms-full.txt is the full content dump: the complete text of your documentation flattened into one giant markdown file, so a model can ingest everything at once without following any links.

Which one you need depends entirely on what kind of site you run:

• Marketing or service business site: llms.txt alone is plenty. An llms-full.txt of your blog adds bulk without a clear consumer.
• Developer product with extensive docs: use both. The llms-full.txt is where the real utility is, because AI coding agents helping developers integrate your product can load your entire reference in one request. This is exactly the pattern Anthropic, Stripe, and Mintlify follow.

One caution on llms-full.txt: for large documentation sets the file can grow well past what a model’s context window can actually hold, at which point the tooling reading it has to chunk it anyway. Keep it scoped to the docs that matter rather than exporting everything you’ve ever published.

Should You Add One? A 30-Second Decision Guide

What Google, OpenAI, and Anthropic Actually Say

This is the part most llms.txt guides skip. Google updated its Search Central documentation in June 2026 under the heading “Clarifying guidance on llms.txt files,” stating explicitly that these files are not needed to appear in Google Search, including its generative AI features such as AI Overviews and AI Mode, and that having or not having one creates neither a positive nor a negative visibility effect. Google’s Gary Illyes had already said as much at Search Central Live in July 2025, confirming Google does not support llms.txt and has no plans to, and that ranking in AI Overviews requires nothing beyond standard SEO fundamentals covered in our SEO guide.

OpenAI, Anthropic, and Meta have not made an equivalent public statement either way as of this writing. In practice, that ambiguity is doing a lot of work in llms.txt marketing copy elsewhere on the web. The absence of a denial is not the same as a confirmation, and the crawl log evidence below points toward the more skeptical reading.

How to Write One (6 Steps)

1. Pick your highest value pages. Don’t dump your entire sitemap in. That defeats the purpose. Choose the pages that best represent what you do: core service pages, your best guides, an FAQ, key policies.

2. Format it in markdown. Use # for your site or company name at the top, a one line > description below it, ## for section headers (Services, Case Studies, Resources), and - [Page title](https://url): short description for each link.

Here’s a complete llms.txt example you can copy and adapt:

# Acme Plumbing
> Licensed plumbing company serving Tampa Bay since 2005. Emergency repairs, water heaters, and repiping.

## Services
- [Emergency Plumbing](https://acmeplumbing.com/emergency/): 24/7 emergency repair, average 45-minute response
- [Water Heater Installation](https://acmeplumbing.com/water-heaters/): Tank and tankless, all major brands

## Resources
- [Pricing Guide](https://acmeplumbing.com/pricing/): Flat-rate pricing for common jobs
- [FAQ](https://acmeplumbing.com/faq/): Answers to the questions customers ask most

## Company
- [About](https://acmeplumbing.com/about/): Licensing, insurance, and service area

That’s the entire format. One H1, a blockquote description, H2 sections, and links with one-line descriptions. Swap in your own pages and you’re done.

3. Upload it to your root directory. It needs to live at yoursite.com/llms.txt, not buried in a subfolder, unless it’s specifically scoped to a docs subdomain. Serve it with a text/plain content type and confirm it returns a clean 200 status with no HTML or PHP errors mixed into the output.

4. Test it. Visit the URL directly in a browser. If it loads as plain, readable markdown, you’re set.

5. Keep it fresh. This isn’t a set it and forget it file. Review it every few months, dropping outdated pages and adding new ones worth spotlighting.

6. Consider an llms-full.txt for documentation heavy sites. If you run a product with extensive docs, a companion llms-full.txt containing the complete text in one file gives a model everything at once, rather than a list of links it may or may not follow. This matters more for software documentation than for a typical marketing site.

Who Is Actually Using It

Adoption skews heavily toward developer tools and technical infrastructure companies, which makes sense given the standard originated in that world. Confirmed adopters include Anthropic, Stripe, Cursor, Cloudflare, Vercel, Mintlify, and Supabase. Shopify is the most aggressive mainstream adopter: it pushed llms.txt to every store on its platform by default in spring 2026, putting merchant adoption above 78% almost overnight, far higher than any sector level figure achieved organically. That single decision by one platform accounts for a meaningful share of the overall adoption growth described above, which is worth keeping in mind before treating the raw site count as evidence of broad, deliberate adoption.

Does It Actually Work?

The evidence so far is thin, and it has gotten more specific, not more encouraging, as more people have actually checked. A widely cited crawl log analysis covering more than 137,000 domains found that 97% of published llms.txt files receive zero requests from any AI crawler. A separate, smaller audit of 1,000 domains found no requests at all from GPTBot, ClaudeBot, or PerplexityBot specifically, none of the three major AI crawlers showed up even once. Oddly, the most cited domains in AI answers actually show lower llms.txt adoption than lower ranked domains, which is the opposite of what you’d expect if the file genuinely influenced citation behavior.

That said, companies like Anthropic, Stripe, and Hugging Face maintain one on their own sites, most likely for developer and coding agent tooling (an AI assistant helping a developer integrate your API benefits from a clean reference file) rather than as a search citation signal. That distinction matters: llms.txt may be genuinely useful for agentic coding tools reading your documentation, while doing essentially nothing for whether ChatGPT or Perplexity cites your homepage in a search answer. Treat it as a low risk experiment scoped to the right use case, not a proven AI search lever. The things that actually move AI citations are covered in our guide on how to get cited by ChatGPT, Claude, and Perplexity.

Tools and WordPress Plugins

You don’t need to hand write the file if your stack already has a generator. On WordPress, All in One SEO (AIOSEO) includes an llms.txt generator that reads your existing content and SEO settings. The dedicated Basis LLMs.txt File Generator plugin creates and maintains the file automatically as your content changes, respecting existing noindex and nofollow settings from Yoast or Rank Math. For a lighter, fully customizable option, the open source WP-Autoplugin llms-txt-for-wp plugin is worth a look. Outside WordPress, most documentation platforms built for developer tools, including Mintlify, now generate llms.txt and llms-full.txt automatically as part of their standard build.

Whichever route you take, validate the output the same way: visit the live URL directly, confirm it returns plain text with no template wrapper or theme header bleeding into it, and check that every link resolves to a real, absolute URL.

Should You Add One?

If it’s quick to set up, add it. It costs almost nothing and can’t hurt, and it may genuinely help AI coding agents and documentation tools that read your site. Just don’t treat it as a substitute for the things that actually move AI citations and Google rankings: structured data (schema markup, see our FAQ schema markup guide), genuinely helpful and well organized content, and real mentions and backlinks (entity SEO territory). That’s the foundation we build into every Winsignia site, with an llms.txt file as a small bonus on top, not the main event.

We practice this on our own site: winsignia.io/llms.txt is live right now, alongside Organization schema and FAQ schema on our homepage. If you want a technical read on where your own site stands, that’s the first step in our process, or go straight to the authority work llms.txt alone can’t replace.

FAQ

Does llms.txt help Google rankings?
No. Google’s own June 2026 Search Central guidance states plainly that llms.txt files are not needed for Google Search, including AI Overviews and AI Mode, and create neither a positive nor a negative visibility effect.

Do ChatGPT, Claude, or Perplexity actually read llms.txt?
There is no public confirmation from OpenAI, Anthropic, or Perplexity that their crawlers use it, and crawl log studies covering over 137,000 domains found 97% of published files get zero AI crawler requests, with GPTBot, ClaudeBot, and PerplexityBot specifically absent from a separate 1,000 domain audit.

What is the difference between llms.txt and robots.txt?
Robots.txt controls access; it tells crawlers what they are and are not allowed to request, and virtually every major crawler honors it. llms.txt does not control access at all. It only suggests which pages a model should prioritize if it happens to read the file, and no honoring behavior is confirmed.

What is llms-full.txt?
A companion file to llms.txt that contains your complete documentation content in a single file, rather than a list of links. It is most useful for software documentation sites and less relevant for a typical marketing or service business site.

Where does the llms.txt file go on my website?
At the root of your domain: yoursite.com/llms.txt, the same place robots.txt lives. It should be served as plain text and return a clean 200 status. If your documentation lives on a subdomain like docs.yoursite.com, a second file scoped to that subdomain is fine.

Who is actually using llms.txt?
Adoption concentrates in developer tools and technical infrastructure: Anthropic, Stripe, Cursor, Cloudflare, Vercel, Mintlify, and Supabase all maintain one. Shopify pushed it to every store on its platform by default in spring 2026, pushing merchant adoption above 78%.

Is llms.txt worth the effort in 2026?
It costs very little to add and cannot hurt, so it is a reasonable low priority addition, especially if you run developer documentation that AI coding agents might reference. It is not a substitute for schema markup, content quality, and third party authority, which are the signals with actual evidence behind them.

Related Reading

Go deeper on GEO and technical foundations:

• What Is GEO? The Complete Guide
• Entity SEO: Build the Brand Authority AI Models Trust
• Structured Data & Schema Markup: The Complete Guide
• Get Cited by ChatGPT, Claude & Perplexity
• ChatGPT for GEO: Stats & What Makes It Different
• AI Search Statistics 2026
• GEO Statistics 2026: What Gets Brands Cited by AI Engines
• GEO Agency: What to Look For in 2026