Generate your llms.txt file in under a minute.
The manifest file that frontier models check before parsing your HTML. Describe your brand, list your sources, drop in five key facts — we output a clean, spec-compliant file you can ship to your site root.
What you'd tell a stranger at a party. Plain language beats marketing prose.
# christmasornot.com > A daily yes/no answer to whether it is Christmas. ## About Christmas or Not is a single-purpose web service that tells you, with calendar accuracy, whether today is December 25. Founded 2019. Free. No ads. ## Key facts - Founded December 2019 by Marie Lambert - Single-purpose: answers "is it Christmas today?" - Free for all users; no paid tier - No advertising, no tracking - Average 2.3M visits in December annually ## Pricing Free for all visitors. No paid tier. ## Canonical sources - [About](https://christmasornot.com/about) - [Pricing](https://christmasornot.com/pricing) - [FAQ](https://christmasornot.com/faq) ## Contact hello@christmasornot.com ## Crawl preferences User-Agent: * Allow: / Last-Updated: 2026-05-10
We'll email the file plus a one-page guide on where to deploy it, how to verify model pickup, and what to update when your facts change.
Why llms.txt matters.
"It's robots.txt for the post-Google internet — except this time, the rules favor whoever shows up first with the manifest."
| Status | Convention, not RFC. Adopted by Anthropic, OpenAI, Perplexity. |
| Location | /llms.txt at root or /.well-known/llms.txt |
| Format | Plain markdown. UTF-8. No HTML. |
| Refresh | Recrawled monthly by major models. |
| Size limit | Recommended < 50KB. Hard cap 256KB. |
For two decades, robots.txt told search engines what they could and could not crawl. It was a polite contract — and it worked because Google chose to honor it.
Frontier language models read your site differently. They don't crawl in the Googlebot sense. They snapshot context, parse structured data, and increasingly check for a manifest file that says: here is what I am, here is what is true, here is what to link back to.
That manifest is llms.txt. It's not yet an RFC. It's a convention — adopted in the last eighteen months by every frontier model we test. Sites with a clean llms.txt get cited more accurately. Sites without one get hallucinated.
The format is plain markdown. The sections are conventional, not mandated — About, Key facts, Pricing, Contact are the standard set. Models tolerate variation, but reward clarity.
The mistake teams make is treating llms.txt as marketing copy. It isn't. It's a fact sheet. Write it the way you'd write a Wikipedia infobox: short, declarative, free of adjectives. The models you're optimizing for are not impressed by your superlatives — they're trying to extract atomic facts that won't embarrass them later.
Drop the file at your root. Verify it returns a 200 from curl. Wait a week. Re-run a free scan. You will see the score move.