Being found in AI answers

Language models take answers from sources that answer one question completely, in clear language, with facts that can be checked. Marketing language without substance yields nothing to extract; there is nothing to quote.

Structure helps: a heading that poses the question, an answer of a few sentences beneath it, and only then the nuance. That happens to be exactly how people scan too.

What works against you: information that exists only in an image or a table, text that appears only after a click, and pages covering three topics at once.

Put questions and answers on the page where they belong and mark them up as question and answer in the structured data. Name concrete numbers, terms and conditions instead of 'sharp prices'. Repeat the core facts in ordinary sentences.

Keep your business details identical everywhere — site, business profile, directories — so a model does not have to choose which version is right. Contradictory details are a reason not to mention you.

Receive machines hospitably: do not block AI providers' crawlers if you want to appear in their answers, make sure your site is readable without scripts, and optionally publish an llms.txt with a short, factual summary of who you are and what you do.

There is no setting that makes a model mention you, just as nobody can guarantee a first place in Google. Anyone promising otherwise is selling something else than they say.

What you can do is be the source that is easiest to cite: clear, complete and verifiable. That is the same investment that convinces people, so it is never wasted.

Measure the effect with what you have: enquiries in which people say they 'read somewhere', traffic without a clear source, and whether your own name comes up when you ask the well-known assistants the questions yourself.