Do ChatGPT, Claude, and Perplexity actually read llms.txt?

No AI engine has made an official public commitment to reading llms.txt as a first-class signal. However, Claude (Anthropic), Perplexity, and retrieval pipelines used by AI tools can and do fetch it when crawling domains — and Anthropic itself publishes one at anthropic.com/llms.txt, which signals it's taken seriously internally. Think of it like robots.txt in 1996: not a formal standard, but increasingly expected and acted on.

What is the difference between llms.txt and llms-full.txt?

llms.txt is a lightweight navigation index — typically under 10KB — with one-sentence descriptions and URLs for each key page. llms-full.txt is a single file containing the complete text of your entire site for AI systems that want to ingest everything at once. Most small businesses only need llms.txt. Only add llms-full.txt if you have dense documentation or reference content — a large guide library, API docs, or comprehensive how-to corpus — that an AI assistant would benefit from reading in full. See the full guide: How to Write llms-full.txt for Your Small Business Site.

Is llms.txt a formal web standard like robots.txt?

Not yet. Jeremy Howard proposed it in September 2024 as an informal open standard, modeled on robots.txt. It has no W3C or IETF backing as of mid-2026, though W3C has discussed it. Adoption is real — about 10% of the ~300,000 domains in SE Ranking's early-2026 study published one, with Anthropic, Stripe, Cloudflare, and Vercel among early adopters — but it's still a voluntary convention, not an enforced protocol.

How long should my llms.txt be?

Keep it under 10KB for the navigation index. A well-structured llms.txt covers your site name, a 3-5 sentence summary, and organized links to your 20-50 most important pages, each with a one-sentence description. Avoid dumping your entire sitemap — AI engines benefit from curation, not volume. If you have hundreds of pages, organize them into meaningful H2 sections and only include the pages that would actually answer the questions your audience has.

Will adding llms.txt hurt my traditional SEO?

No. llms.txt sits at a static URL (yourdomain.com/llms.txt), is plain text, and is not indexed by Google as a ranking-relevant page. It does not interfere with your robots.txt, sitemap.xml, or any other SEO signal. The only risk is writing an inaccurate llms.txt that points AI engines to outdated or wrong information — keep it updated when you restructure your site.

llms.txt: What It Is, Why It Matters, and How to Write One

TL;DR

llms.txt is a plain-text markdown file placed at yourdomain.com/llms.txt. It tells AI systems — ChatGPT, Claude, Perplexity, Google AI Overviews — what your site is about and where to find your best content. It was proposed in September 2024 by Jeremy Howard (Answer.AI / fast.ai) and has since been adopted by names like Anthropic, Stripe, Zapier, Cloudflare, and Vercel — roughly 10% of the 300,000 domains in SE Ranking’s early-2026 study. No AI engine officially requires it, but it functions exactly like robots.txt did in the early web: a voluntary, lightweight convention that helps crawlers work faster and more accurately.

What Is llms.txt?

When an AI assistant like Perplexity or Claude fetches your website to answer a user’s question, it reads your HTML the same way a browser would — except it doesn’t see your layout, navigation, or visual hierarchy. It sees a wall of text.

llms.txt is a remedy. It’s a curated, machine-readable index of your site in clean Markdown: no HTML tags, no ads, no cookie banners. Just your site’s purpose and your most important pages with one-sentence descriptions of each.

The proposal came from Jeremy Howard, who published it at llmstxt.org on September 3, 2024. Howard modeled it on robots.txt — a plain-text file that every web developer already knows how to create and that crawlers already know to look for.

The format spec is intentionally minimal:

Element	Required?	What it does
`# Site Name`	Yes	Identifies the site for AI context
`> Blockquote summary`	Recommended	2-4 sentence plain-English description
`## Section headings`	Optional	Organize pages into logical groups
`[Link](URL) — description`	Optional	Points to key pages with one-liner context
`## Optional` (special section)	Optional	Secondary content; AI can skip to save context

Who’s Actually Using It?

The most reliable adoption data comes from a SE Ranking study of roughly 300,000 domains in early 2026: about 10% had published an llms.txt file. Adoption was remarkably even across traffic tiers — around 9.9% for low-traffic sites and 8.3% for the highest-traffic sites — but it skews heavily toward developer-tool companies and tech-forward content sites. Among the notable adopters:

Anthropic — anthropic.com/llms.txt
Stripe — developer documentation
Zapier — workflow documentation
Cloudflare — developer docs
Vercel — developer docs
SEOPulse — seopulse.ai/llms.txt

The flip side: that same study found llms.txt files clustered in tech and SaaS, while compliance-heavy sectors — financial services, healthcare, legal — published them on under 10% of top-100 domains.

The practical conclusion: llms.txt is no longer niche, but most competitors in most verticals still haven’t added one. That gap is an opportunity.

Does It Actually Work?

Here’s the honest answer: no major AI platform has made a public, documented commitment to reading llms.txt as a first-class signal. There’s no official API documentation from OpenAI or Google saying “we read this file.”

What is confirmed:

Claude fetches and uses llms.txt when crawling domains (Anthropic publishes one, which signals internal adoption)
Perplexity’s crawler reads llms.txt when present
Several AI retrieval pipelines use llms.txt to prioritize which pages to fetch when context windows are limited

The analogy to robots.txt holds: no browser or crawler was required to follow robots.txt in 1994. They did anyway because it was useful. llms.txt is following the same adoption curve.

The lowest-risk way to think about it: it takes 30 minutes to create and zero ongoing maintenance if your site structure doesn’t change. The upside — better AI comprehension of your content — is real even if the exact weighting is unknown.

The llms.txt Template

Here’s a copy-paste template for a small business or content site:

# Your Business Name

> One paragraph (2-4 sentences) describing what your business is, who you serve,
> and what problems you solve. Write this as if you're explaining to someone
> who has never heard of you. Be specific: include your location if local,
> your product category, and what makes you different.

## Services / Products

- [Service Page Name](https://yourdomain.com/services/page) — What this service is and who it's for
- [Product Page](https://yourdomain.com/product) — What the product does, price range, who buys it

## Key Guides & Resources

- [Blog Post Title](https://yourdomain.com/blog/post-slug) — One sentence on what question this answers
- [How-To Guide](https://yourdomain.com/blog/guide) — What task this helps the reader complete

## About

- [About Page](https://yourdomain.com/about) — Founded year, location, mission, key team
- [Contact / Locations](https://yourdomain.com/contact) — Hours, phone, address

## Optional

- [Older resource](https://yourdomain.com/blog/older-post) — Still accurate but not a priority

What to avoid:

Dumping your entire sitemap — curation matters more than volume
Vague descriptions (“our best content”) — be specific about what the page answers
Links to thin or low-quality pages — you’re telling AI engines what to trust
Letting it go stale — update it when you remove or restructure pages

llms.txt vs. llms-full.txt

The spec also defines a sibling file: llms-full.txt. The difference:

	llms.txt	llms-full.txt
Content	Index of links with descriptions	Full text of every page
Size	Under 10KB	Can be megabytes
Use case	Navigation — AI decides which pages to fetch	Bulk ingest — AI reads everything at once
Who needs it	Most sites	Documentation-heavy sites, developer tools

For most small businesses and content blogs, llms.txt alone is sufficient. Add llms-full.txt only if you have dense reference content — API docs, legal guides, comprehensive how-to libraries — that an AI assistant would benefit from reading in full.

Adding llms.txt to Your Site

Static sites (Astro, Next.js, Hugo): Place the file in your public/ directory. It will be served as-is at /llms.txt.

WordPress: Place it in your theme’s root directory or upload it via FTP to your domain’s public_html folder. Use a redirect plugin if needed to serve it from the root.

Verify it’s live: After deploying, check https://yourdomain.com/llms.txt in your browser. You should see plain Markdown text, not HTML. The response header should show Content-Type: text/plain.

Once live, run your domain through the free SEOPulse audit — it checks for llms.txt presence as part of the GEO readiness score, alongside schema markup, FAQ structure, and answer-first content signals.

Where llms.txt Fits in Your GEO Strategy

llms.txt is a good first step, but it’s one signal among several. The bigger GEO wins come from content structure: answering the question directly in the first paragraph, using question-based headings, publishing specific verifiable facts, and keeping your factual claims consistent across your own pages and external sources.

For the full picture, read the companion guides: