llms.txt: the one file AI crawlers actually read
Search engines read robots.txt and sitemap.xml. AI assistants increasingly read a third file: llms.txt.
What llms.txt is
llms.txt is a plain-text file at your site’s root that describes your site for language models. Where a sitemap lists URLs, llms.txt explains what your site is about, which pages matter, and what a model should read to answer questions about you accurately.
In practice it’s a short, curated list: a one-line description of your company, followed by your most important pages and what’s on them, like a README for AI crawlers.
Why it matters for GEO
AI crawlers are impatient. If your key claims, statistics, and definitions are buried across a hundred pages, the model may cite a competitor who made it easy. llms.txt gives the crawler a curated map, and models that read it get your story right the first time.
What belongs in it
- A one-line description of who you are and what you do.
- Your most citable pages: product, pricing, comparisons, definitions.
- Your strongest claims and stats: the exact numbers you want quoted.
- Links to structured data and docs the model can pull facts from.
What not to do
- Don’t paste your whole site. Curate. A 200-line file beats a 20,000-line dump.
- Don’t stuff keywords. Write for a model that needs to understand you, not a ranking algorithm.
- Don’t set it and forget it, keep it in sync when your product or pricing changes.
The takeaway
llms.txt is one file, but it’s the difference between “the model gets my brand right” and “the model cites my competitor.” If you’re doing GEO, this is step one.