Answer engine

An answer engine responds with an answer instead of a list of links. ChatGPT Search, Perplexity, Google AI Overviews and Claude’s search all behave this way.

How does this differ from a search engine?

A search engine returns candidates and lets the user choose, which sends traffic to the chosen site. An answer engine reads the candidates itself and returns a synthesis, which often satisfies the question without any onward click.

Why does that break publisher economics?

The content still does the work of answering, but the visit that funded it no longer arrives. The exchange that supported the open web, content in and traffic back, quietly stops while the reading continues.

Can a publisher still benefit from being read?

Yes, in two ways. Being cited builds authority and occasionally returns a visitor. And the crawl itself can be monetized directly, which is what CrawlClick does — so the read returns value whether or not a visit follows.

Do all answer engines work the same way?

No, and the differences decide which of them you can influence. Some run their own index and crawl it in advance; some fetch pages live at the moment of the question; several do both with separate agents and separate tokens. Some show numbered citations the reader can follow, some name a source in passing without a link, and some return a paragraph with no attribution at all. The engine that links is the only one that can send you a visitor, and that is a property of the product rather than of your page.

Why does the same question give different answers?

Because generation is not deterministic and retrieval is not stable. Ask twice and the set of retrieved passages can differ, the ordering can differ, and the wording certainly will. This is why a single observation proves nothing about whether you appear in answers, and why any measurement worth having is a repeated sample rather than a spot check.

What can a publisher actually see from the outside?

Very little without asking. No answer engine publishes how often an answer citing you was shown, how many people read it, or what fraction of them went on to click anything. There is no impressions log and no equivalent of a rank report supplied by the vendor. Everything a publisher knows about their presence in answers has to be sampled by running the questions themselves, or inferred from the traffic that does arrive.