Riccardo Terzi leads Google’s News Partnerships. His latest piece on how online search is changing is among the clearest things I have read this year. It is worth explaining why.
For decades, search worked one way. Type in keywords, scroll through a list of links, open page after page until you found the answer. Terzi cites a metric I have used myself for years to judge the quality of a site, pogo sticking. It is when a user clicks a result, does not find what they were looking for, and immediately hits the back button to try elsewhere. Google pays close attention to this behavior. The reason is simple: it thinks of itself as an answer engine first, a search engine second. If the user bounces back, something in the content failed.
Today that model is breaking down. ChatGPT, Perplexity, and Google SGE no longer just show links. They answer, summarize, and steer the conversation inside an interface that feels like an assistant rather than a list of results.
AI, in this shift, decides what gets shown and how. It is a gatekeeper of information. Users no longer need to click a link or visit a site: the answer arrives ready-made, inside the conversational environment.
The consequence is that web traffic loses weight as a KPI for publishers and marketers. In the old SEO model, traffic was everything: more visitors, more ad revenue. If the answer arrives with no click at all, that equation stops holding.
This is where GEO comes from, Generative Engine Optimization. It no longer aims to climb a ranking. It aims to get chosen by generative algorithms when they build a direct answer for the user. Getting there takes real editorial quality, semantic depth, and a point of view that cannot be found anywhere else.
I already wrote about this in From Brand to Source, an earlier piece on the same shift.
To become AI-friendly, understandable and usable by artificial intelligence, content needs structured data and advanced semantic markup. And you need to monitor your real presence inside the answers AI assistants actually generate, not the one you imagine you have.
Tools to measure this are emerging too. Relixir, for one, is a Y Combinator-backed startup building analytics dedicated to visibility in generative results. Then there are the new technical standards, NLWeb and MCP, pushed by Microsoft and Anthropic, which let AI agents interact directly with content.
And as traffic, measured the way we have always measured it, declines, new revenue models are emerging alongside it. Licensing for content used by AI algorithms. Micro-transactions. Premium content distributed via API. Plus direct loyalty strategies, communities, newsletters, exclusive services, that do not depend on organic traffic at all.
I wrote about this elsewhere too, in Toward a sustainable digital publishing model.
Terzi’s piece confirms a warning bell that had already rung elsewhere. But it is also an opportunity for whoever moves before everyone else. The shift from SEO to GEO changes teams, priorities, the way you measure results. It is a cultural change before it is a technical one. Waiting to decide whether to adapt already means losing ground. The real question is how to do it, and where to start.
Frequently asked questions
