# Which facts should a machine be able to read about your business? Topic: [Machine-readable website](https://btlabs.dev/en/ai-website) > Not whether to add structured data, but which: five facts that decide whether an AI can recommend you — plus the two-minute check that fails most often. A machine that reads your website has to decide, in a fraction of a second, whether it can answer a question with what it finds. Most sites make that decision easy in the wrong direction: they describe everything and declare nothing. The question is not whether to add structured data. It is **which facts** — and in which order. ## Start with the facts that cost you money when they are wrong. Not everything on your site deserves markup. A useful test: if this detail were wrong in an AI answer, what would it cost? - **How to reach you.** Phone, email, the form. If an assistant cannot pass on a phone number, it may well name someone whose number it can read. - **When you are open.** Including the exceptions — holidays, seasonal hours, the winter emergency service. - **Where you work.** Address is not the same as service area. A business that serves three valleys should say so. - **What you do — and what you expressly do not.** The second half is the one everybody skips, and it is what stops you being recommended for the wrong job. - **Roughly what it costs.** A range, a starting price, an hourly rate. Anything is better than an empty field. Everything else — your history, your values, the team photo — is for human readers. That is not a criticism; it is a division of labour. ## The order matters more than the completeness. A common mistake is to mark up everything at once and maintain none of it. Three facts that are correct beat thirty that were correct last year. Start with the ones above, in that order. They are also the ones people actually ask assistants about, which is why they are the ones worth keeping current. ## The check nobody runs Here is the test that fails most often, and it takes two minutes: **open one of your pages and compare what the markup says with what the visible text says.** We have seen sites serving 2023 opening hours in machine-readable form while the visible text had the current ones. To a human reader the page is fine. To a machine it is wrong — and the machine believes the marked-up value, because that is the one that was declared deliberately. This is not a rare edge case. It is what happens whenever markup is written once and the text is edited afterwards. Which is to say: always, eventually. ## Declared once, or maintained in five places That leads to the decision that actually matters, and it is not a technical one. If your structured data is written by hand, it is a copy — and copies age. If it is generated from the same source as the visible page, it cannot drift, because there is nothing to drift from. Change your opening hours once and every output changes with them: the visible text, the markup, the summary an AI agent reads first. The difference does not show on the day of setup. It shows the first time something changes. ## What to take away Ask three questions about your own site: 1. Is there any structured data at all? (If you cannot say, the answer is usually no.) 2. Does it cover the five facts above — or the ones that were easy to add? 3. If a detail changes tomorrow, how many places do you have to touch? If the answer to the third is more than one, one of them will eventually be wrong. Not perhaps — certainly. What structured data is and how the vocabulary works is explained in [Structured data](https://btlabs.dev/en/glossary/structured-data). Which **type** of business you declare — the question that comes before all of this — is in [What are you, to a machine?](https://btlabs.dev/en/posts/what-are-you-to-a-machine). And who is accountable when a machine gets it wrong anyway is in [When the machine answers for you](https://btlabs.dev/en/posts/when-the-machine-answers-for-you). --- Source: https://btlabs.dev/en/posts/what-structured-data-to-add Last-Modified: 2026-09-30T07:00:01.514Z Languages: [de](https://btlabs.dev/llms/de/posts/welche-daten-auszeichnen) · [it](https://btlabs.dev/llms/it/posts/quali-dati-strutturati-inserire) See also: [llms.txt](https://btlabs.dev/llms.txt) · [ai.txt (Policy)](https://btlabs.dev/ai.txt) · [identity.json](https://btlabs.dev/identity.json)