Answer engine

An answer engine responds to a question with a composed answer instead of a list of links. ChatGPT with web search, Perplexity and Google AI Overviews all work this way. Search still finds pages, but what gets shown is the result of reading them.

How does an answer engine work?

In three steps. The question is broken into sub-questions, sources are retrieved for each, and an answer is assembled from the passages found. That answer is newly written and does not come verbatim from any single source.

An unfamiliar consequence follows for visibility: the competition happens at section level. What gets judged is not whether your page is good overall, but whether one particular paragraph answers one particular sub-question better than other paragraphs on the web.

Technically an answer engine is therefore close to RAG: the same construction, with the web as the document corpus instead of your files.

What changes for marketing

Three things.

Questions get longer. Instead of "CRM mid-market", people ask "which CRM suits a 30-person trades business with no IT department". Questions like that carry the use case, and general pages cannot answer them.

The question comes earlier. People use these tools while mapping a market, before they compare vendors. Not being named at that stage means not being on the list later.

Measurement shifts. An answer engine sends fewer clicks because many questions are fully answered inside the tool. The effect comes through mention, see zero-click search.

How to appear there

As described under Generative Engine Optimization, with one addition: answer the questions your customers actually ask, in the wording they ask them.

A glossary entry like this one is a good format for that, because it names a question and answers it directly. A services page that opens with company history is not.

You can check this without any tool: ask an answer engine the ten questions customers usually start with, and note who gets named. That result is the baseline against which any visibility work gets measured.

Which answer engines matter

Four surfaces are relevant for the German-speaking market, and they behave differently.

Google with AI Overviews has the widest reach, because the answer appears inside the search people already use and nobody has to adopt a new tool.

ChatGPT with web search is where a lot of research now begins. Answers are longer and the number of cited sources is small, which sharpens the competition for a mention.

Perplexity leans harder on citation and shows its sources more prominently. Reach is smaller, but the audience is often specialist.

And the assistants built into other tools, in browsers or office applications, are growing in importance without being perceived as search at all.

In practice that means the baseline should be taken on at least two surfaces. Check only one and it is easy to mistake a quirk of that surface for a finding about your own visibility.