When nobody googles any more: visibility in AI answers

Search has changed more in a short time than in the fifteen years before. More and more people put their question to an assistant and get an answer – with no results list. For companies that means: it is no longer enough to be in first place. You have to appear in the answer.

Why classic search engine optimisation is not enough

An AI assistant does not read meta titles and does not count links in the old sense. It looks for passages that answer a question cleanly, and for sources it rates as reliable. Anyone who describes their services in advertising language provides nothing that can be quoted.

What helps: answers instead of praise

On each of our service pages there is a block "In short" – three to five sentences that directly answer the most common question, with place, figures and limitations. Exactly such paragraphs are what assistants pick up. Add real questions and answers at the bottom of the page, marked up as FAQ so that machines recognise the structure.

Technical prerequisites

Measure instead of hope

Every month we ask the big assistants the same questions and log whether and where a company is mentioned. Only then does it become visible whether a measure works. Without measurement, visibility in AI answers is a gut feeling.

What we actually implement

For every service page we write an answer block of three to five sentences that answers the most common question, naming place, figures and limits. Below that are real questions from client conversations with answers of two to four sentences. Both are marked up as structured data.

Add a file llms.txt in the root directory: a machine-readable summary of the company with facts, services and addresses. It is not an official standard, but several assistants read it and it costs little effort.

The difference between AEO and GEO

Both terms are in circulation. Answer Engine Optimization means optimising for answer engines in general, Generative Engine Optimization specifically for generative systems. In practice they overlap heavily: clear answers, verifiable facts, clean structure, recognisable authorship.

What does not work

Keyword lists in body text, hidden text, artificially generated mass text. Assistants assess coherence and agreement with other sources. Anyone who claims to be market leader without it being stated anywhere else does not get quoted.

How often to measure

Monthly is enough. We ask the big assistants the same ten to twenty questions that potential clients would ask and log whether and how the company appears. The development over six months shows whether a measure is working.