emerging convention — honest assessment
llms.txt: what it is, and whether AI actually reads it
Last updated
Plenty of GEO advice presents llms.txt as a must-have. The evidence says otherwise, and honesty about evidence is this wiki’s whole premise — so here is what llms.txt is, what the adoption data shows, and the narrow cases where it is still worth ten minutes.
What is llms.txt?
llms.txt is a proposed convention: a markdown file at your site root (/llms.txt) giving language models a curated map of your most important pages with one-line descriptions. It was proposed in September 2024 by Jeremy Howard of Answer.AI and is documented at llmstxt.org.
The idea is reasonable on its face: context windows are finite, HTML is noisy, so hand the model a clean index. The open question was never elegance — it was whether any AI system would actually fetch the file.
Do AI crawlers actually read llms.txt?
Mostly no. No major LLM provider — OpenAI, Anthropic, Google, or Meta — has formally adopted the convention, and an Ahrefs analysis of 137,000 domains found that while about 28% now publish an llms.txt file, 97% of those files received zero requests in May 2026.
Google’s John Mueller has compared it to the keywords meta tag — self-reported metadata that systems learned to ignore — while allowing it may serve as “a temporary crutch” for niche cases like AI coding tools parsing developer documentation.
Should you publish an llms.txt anyway?
If it costs you ten minutes and you will keep it current — sure; the downside is zero and the option value is real if adoption arrives. But it belongs at the bottom of your GEO list, after every fix with measured impact today.
- Reasonable: documentation-heavy products, sites already generating it from a sitemap, teams that automate freshness.
- Waste of focus: treating it as the fix while retrieval bots are blocked or your prices live in a PDF.
- Actively bad: a stale llms.txt confidently describing pages you deleted last year.
This mirrors how the GEOExtension audit scores it: an “emerging” rule worth 1 point out of roughly 60, flagged as low priority — never a reason a site fails.
How do you write an llms.txt file?
One H1 with your name, a short blockquote summary, then sections of links with one-line descriptions — plain markdown, most important pages first, nothing else. Here is a complete example for a local business:
# Riverside Dental Care > Family dental practice in Boise, Idaho. Preventive care, > implants, and same-day emergency appointments. ## Services - [Preventive care](https://example.com/services/preventive): cleanings, exams, x-rays - [Dental implants](https://example.com/services/implants): pricing and process - [Emergency dentistry](https://example.com/emergency): same-day appointments ## About - [Our dentists](https://example.com/team): credentials and bios - [Insurance & pricing](https://example.com/pricing): accepted plans, cash prices - [FAQ](https://example.com/faq): common patient questions
Frequently asked questions
Is llms.txt an official standard?
No. It is a community proposal from September 2024 (Jeremy Howard, Answer.AI) hosted at llmstxt.org. No standards body has ratified it, and no major AI provider has committed to reading it.
Do OpenAI, Anthropic, Google, or Perplexity use llms.txt?
As of mid-2026, none has formally adopted it. Ahrefs’ log analysis found 97% of published llms.txt files received zero requests in May 2026 — the clearest available evidence that AI crawlers are not fetching it at scale.
Will llms.txt hurt my site if I add one?
No. It is a plain markdown file that nothing else reads; the realistic downside is only the maintenance cost of keeping it in sync so it never serves stale facts if adoption arrives.
Is llms.txt the same as robots.txt?
No — they are opposites in spirit. robots.txt is an enforced convention that restricts crawlers; llms.txt is an unadopted convention that offers crawlers a curated content map. Only robots.txt currently changes bot behavior.
Why does the GEOExtension audit still check for llms.txt?
As an explicitly flagged “emerging” check with the lowest weight in the rulebook — worth 1 point of roughly 60. Missing it produces a soft note, never a failing grade, which mirrors the actual evidence for it today.
What should I prioritize instead of llms.txt?
The checks with real observed impact: retrieval-bot access in robots.txt, content readable without JavaScript, valid Organization and FAQPage schema, question-shaped pages with direct answers, and cited statistics. All are covered in the free audit.
see where you stand
Is AI already recommending your business?
Run the free audit to score any page against the 19 GEO checks this wiki teaches — no account, no API keys. Then probe real ChatGPT, Claude, Gemini, and Perplexity answers with your own keys to measure your actual mention rate.