llms.txt is a proposed markdown file placed at the root of a website (yoursite.com/llms.txt) designed to give AI systems a clean, curated map of the site’s most important content. It matters for SEO in the AI era not because it is a magic ranking signal, but because it is confirmed to be read by some AI platforms (Anthropic’s Claude, Perplexity) and explicitly ignored by others (Google Search and AI Overviews). This guide covers what llms.txt actually is, who genuinely uses it, and whether your site should implement it in 2026.
Key takeaways
- llms.txt was proposed by Jeremy Howard of Answer.AI on 3 September 2024. It is a community convention, not a ratified W3C or IETF standard.
- Google has publicly stated it does not use llms.txt. Per Google’s official AI features documentation, updated in 2026, “You don’t need to create new machine readable files, AI text files, markup, or Markdown to appear in Google Search.”
- Anthropic and Perplexity have publicly confirmed support, per the State of llms.txt 2026 research. OpenAI has not officially confirmed ChatGPT support. Mistral reads it with caveats.
- An SE Ranking machine-learning study of 300,000 domains found no correlation between llms.txt presence and AI citation frequency. Adoption sits at roughly 10% of studied domains.
- Implement llms.txt if your audience includes Claude, Perplexity, or developer tools consuming your documentation. Do not implement it expecting Google ranking benefits.
What Is llms.txt?
llms.txt is a markdown-formatted file placed at the root of a website (yoursite.com/llms.txt) that summarizes the site’s most important pages in a format large language models can parse quickly. Think of it as a robots.txt or sitemap.xml for AI systems, though unlike those established standards, llms.txt has no formal endorsement from search engines or standards bodies.
The format is simple. A short site description at the top, followed by curated links to key pages grouped under headings like Docs, Blog, or API Reference. A separate file called llms-full.txt contains the full text of the site’s most important pages, so an AI can retrieve the actual content without having to parse HTML from every page. The proposal originated with Jeremy Howard of Answer.AI in September 2024 and is maintained through the llmstxt.org community project.
Who Actually Uses llms.txt in 2026 (and Who Does Not)?
Support in 2026 splits cleanly. Anthropic and Perplexity have publicly confirmed reading llms.txt. Google has publicly said it does not. OpenAI has not committed either way but shows observable patterns. Mistral reads it. Meta, Microsoft Copilot, and Bing have not announced support.
In practical terms:
- Anthropic (Claude): publicly confirmed. Claude.ai and Claude Desktop respect llms.txt directives in retrieval workflows.
- Perplexity: publicly confirmed. Perplexity retrieves llms.txt and uses it to help prioritize which pages to pull into an answer. (Some analyses dispute the extent, but the file is read.)
- OpenAI (ChatGPT): unconfirmed. Teams publishing llms.txt report correlated shifts in ChatGPT search citation patterns, but this is suggestive, not proven.
- Google (Search, AI Overviews, Gemini): confirmed to NOT use llms.txt. Gary Illyes stated this at Search Central Live in July 2025. John Mueller compared the idea to the deprecated meta keywords tag. Google’s own AI features documentation reinforces this.
- Mistral: reads llms.txt where present, with caveats around spec maturity.
- Developer tools: Cursor, GitHub Copilot, and other coding agents depend on llms.txt for documentation retrieval.
Companies that publish llms.txt today include Anthropic, Cloudflare, Stripe, and Zapier. Adoption across the wider web sits around 10% of studied domains, concentrated in SaaS, developer tooling, and technical publishing.
Does llms.txt Actually Help SEO?
It depends entirely on what you mean by SEO. If SEO means Google organic rankings and Google AI Overviews visibility, llms.txt does not help. Google has stated this directly. If SEO means being cited by Claude, Perplexity, coding agents, and RAG-based enterprise search tools, then llms.txt does help those specific channels.
The evidence on citation impact is worth reading honestly. An SE Ranking machine-learning study of 300,000 domains, cited in OrganiKPI’s 2026 adoption analysis, found no correlation between llms.txt presence and AI citation frequency. When the llms.txt variable was removed from their XGBoost prediction model, accuracy actually improved. That does not mean llms.txt is useless for the platforms that read it. It does mean the “publish llms.txt and get more AI citations” claim is not supported by aggregate data.
In our own SEO and AI-search work at Morphiaas, a performance marketing and creative agency serving India and the US, the honest recommendation is this: implement llms.txt if it costs almost nothing and your audience includes Claude, Perplexity, or developer tools. Do not implement it hoping for Google traffic. And never let vendors sell it to you as a guaranteed AI-visibility shortcut, because that claim has no evidence behind it.
How Do You Create an llms.txt File Correctly?
Seven practical steps produce a well-formed llms.txt file that works across the platforms which actually read it. None of these steps guarantee AI citations, but all of them make sure the file is technically valid and useful for the AI systems that consume it.
1. Structure the file in markdown, not plain text
The llms.txt spec expects markdown formatting: an H1 with your site name at the top, a short blockquote description, an optional details paragraph, then H2 sections grouping curated links (Docs, Blog, API, Company, etc.). Plain-text llms.txt files exist, but the markdown-structured version is what current parsers expect. Follow the llmstxt.org spec.
2. Curate the links carefully, do not dump the full sitemap
The point of llms.txt is a curated map of what matters, not a comprehensive index. Include the pages an AI system would need to answer questions about your brand, products, or content. Skip administrative pages, tag archives, thin content, and duplicate landing pages. A tight file of 30 to 60 curated links outperforms a dump of 500 URLs on both parseability and inclusion decisions.
3. Add descriptions to each link, not just URLs
The spec format for each link is: `[Link Title](URL): short description of what the page covers`. Descriptions help AI systems understand which page to reference for which question. Skip descriptions and you leave the AI to guess based on URL slugs alone, which is a weaker signal than a clear human-written summary.
4. Consider an llms-full.txt alongside the base file
llms-full.txt is a companion file containing the full markdown text of your most important pages inlined into one file. Useful for documentation-heavy sites and RAG use cases because it saves the AI from having to fetch and parse each linked page individually. Skip llms-full.txt for large content sites where the file would balloon past useful size, and skip it entirely if your priority is Google SEO (Google ignores both files).
5. Place the file at the root and confirm it is publicly reachable
The file must live at yoursite.com/llms.txt (and yoursite.com/llms-full.txt if you publish one). Not in a subdirectory. Not behind login. Fetch it in a browser after upload to confirm it returns 200 OK and displays as plain markdown. On WordPress with Rank Math, use a plugin or a small manual upload via FTP or file manager. On Next.js, Astro, or similar frameworks, add it to the public/static folder.
6. Make sure your robots.txt actually allows AI crawlers first
llms.txt is useless if AI crawlers cannot reach your site at all. Check that robots.txt does not block ChatGPT-User, GPTBot, ClaudeBot, PerplexityBot, and Google-Extended. This is the more foundational fix. Sites blocking AI crawlers in robots.txt then publishing llms.txt are optimizing for a door they have locked from the outside.
7. Keep llms.txt updated, treat it like a sitemap
A stale llms.txt is worse than none. If your site publishes new key pages, cover them in the file within a few weeks. Automate regeneration where possible (there are open-source generators for major CMS platforms). Treat llms.txt maintenance the way you treat sitemap.xml maintenance: not glamorous, but the moment it drifts from reality, the file becomes noise.
How Do You Measure Whether Your llms.txt Is Working?
Measurement is genuinely hard because most AI platforms do not publish llms.txt hit logs. The workable approach: check your server logs for retrievals of /llms.txt by AI crawler user agents, run monthly manual audits of your citation presence in Claude, Perplexity, and ChatGPT for your top branded and category queries, and track direct traffic patterns that do not match paid or organic search sources.
A practical stack: your web server access logs (Apache, Nginx, or your hosting control panel) to see whether ClaudeBot, PerplexityBot, or GPTBot are fetching the file, a spreadsheet log of monthly AI-answer audits on your top 20 target queries, and Google Search Console for branded search volume changes that might correlate with AI-driven brand recognition. Judge quarterly. Weekly measurement produces noise on this signal.
If you can state which AI crawlers actually fetched your llms.txt last month and whether your citation presence in Claude or Perplexity changed over the last quarter, your measurement is working. If you are hoping Google ranking will move because you added llms.txt, the measurement will never confirm it because that effect does not exist.
Common Mistakes and Misconceptions About llms.txt
- Believing llms.txt affects Google rankings. Google has publicly stated it does not use the file. Any vendor claiming otherwise is contradicting Google’s own documentation.
- Publishing llms.txt while blocking AI crawlers in robots.txt. The two files work together. If robots.txt blocks GPTBot or ClaudeBot, llms.txt cannot help.
- Dumping the full sitemap into llms.txt. The file is meant to be curated, not comprehensive. A tight 30 to 60 link file outperforms an unfiltered 500-URL dump.
- Skipping the descriptions. URL-only entries force AI systems to guess page content from the slug. Descriptions do most of the interpretive work.
- Treating llms.txt as a formal standard. It is a community convention with no W3C or IETF ratification. Platform support can change without notice.
- Expecting citation lift as automatic. SE Ranking’s 300,000-domain analysis found no correlation between llms.txt presence and citation frequency. The file may help the platforms that read it, but no evidence supports an automatic citation-volume boost.
- Publishing once and forgetting. A stale llms.txt sends AI systems to pages that no longer exist or misses your newest important content. Refresh it on the same rhythm as your sitemap.
- Buying “llms.txt for SEO” as a shortcut. Vendors sell packaged llms.txt generation as an AI-SEO win. Given the evidence, the file is a low-cost hygiene move, not a growth lever. Treat it accordingly.
Build an AI-Search-Ready Site Without the Snake Oil
If you want a technically clean site that shows up honestly in AI search alongside Google, that is the kind of work we do at Morphiaas. We combine technical and content SEO with performant website builds that are crawlable by both Google and AI systems, without overpromising outcomes no vendor can actually guarantee. Book a call and we will audit where your site currently stands on both traditional and AI-search fundamentals.
Frequently Asked Questions
Does llms.txt affect Google search rankings?
No. Google has publicly confirmed it does not use llms.txt. Google’s own AI features documentation, updated in 2026, explicitly states you do not need to create machine-readable AI files or markdown files to appear in Google Search or its AI features.
Which AI platforms actually read llms.txt in 2026?
Anthropic (Claude.ai and Claude Desktop) and Perplexity have publicly confirmed support. Mistral reads it with caveats. OpenAI (ChatGPT) has not officially confirmed but shows some observable response. Google, Meta, Microsoft Copilot, and Bing have not announced support.
Is llms.txt an official web standard?
No. llms.txt is a community proposal introduced by Jeremy Howard of Answer.AI in September 2024, maintained through llmstxt.org. It has not been ratified by the W3C, IETF, or any AI provider as a formal standard. Different platforms may interpret edge cases differently.
Should my website have an llms.txt file in 2026?
Publish it if your audience includes Claude, Perplexity, or developer tools consuming documentation, and if implementation costs are near zero (which they typically are). Skip it if you were hoping to influence Google rankings. Prioritize traditional SEO fundamentals, quality content, and crawlability first.
What is the difference between llms.txt and llms-full.txt?
llms.txt is a curated map of important pages with links and descriptions. llms-full.txt is a companion file containing the full markdown text of those pages inlined into one document, so AI systems can retrieve full content without fetching each page separately. Documentation-heavy sites benefit most from publishing both.