llms.txt

If robots.txt tells crawlers what they may access and sitemap.xml lists your pages, llms.txt does something neither does: it tells an AI model what your site is about, in language the model can read without wading through navigation, ads, and JavaScript.

Where it came from

The convention was proposed in September 2024 by Jeremy Howard, co-founder of Answer.AI and fast.ai, with an open specification hosted at llmstxt.org. It caught on quickly: by 2026, companies including Anthropic, Stripe, Vercel, and Cloudflare publish one. It is still an emerging convention rather than a formal web standard, which is exactly why adopting it early is close to free leverage.

What's in the file

It's plain Markdown. A title and a short summary of what the site or business is, then grouped lists of links with a sentence on what each one covers. The spec also defines a companion llms-full.txt that includes the actual text of key pages, so a model can ingest the content itself rather than just the map.

Why a CRE website should have one

Commercial real estate research increasingly starts inside ChatGPT, Claude, and Perplexity, and those tools synthesize a few sources rather than returning ten links. A clean llms.txt makes it easier for a model to understand who you are, which markets you cover, and what each listing is — and to cite you accurately. It's one piece of a broader picture covered in how CRE sites get surfaced by AI search.

See it in practice

You can read this site's own file at /llms.txt, and the fuller variant at /llms-full.txt. Every MemmoCRE site generates both automatically — no plugin, no contractor. Back to the glossary.