Grounding pages and llms.txt: your website's instruction manual for AI

Imagine an AI system visiting your website for the first time. It has a few seconds to understand who you are, what you offer and whether it should recommend you. A grounding page and an llms.txt are built for exactly this moment.

The grounding page

A grounding page is not a classic landing page. It deliberately avoids advertising language and instead delivers a precise entity factsheet: name, type, location, services, founding date, sources. It’s readable for humans, but built primarily so that AI systems, search engines and knowledge graphs can resolve a company unambiguously.

The most important part is the clean separation of entities. A person is not the same as the company they founded. Mixing the two creates ambiguity — and ambiguity is what makes AI answers unreliable.

The llms.txt

The llms.txt is a lightweight text file in the root directory of your domain — comparable to robots.txt, but for language models. It contains a concise summary of your company and points to the most important pages. A complementary llms-full.txt may be more detailed and deliver all core facts in one place.

The advantage: instead of working through nested HTML, a model finds the essential information immediately and in a machine-friendly form.

What belongs in them

Both formats live on verifiability. Name only facts that are true and ideally backed up externally. Don’t invent profiles that don’t exist — missing sources do less harm than false ones. And keep both up to date: a grounding page with an outdated date loses trust.

Grounding pages and llms.txt are no black magic, but they demand discipline and a clear understanding of your own positioning. That very clarity is the real value — it helps not only machines, but also forces you to say precisely what you stand for.