“Google is dead” feels overstated. What is actually breaking is the click based retrieval and attribution model once answers start getting synthesized upstream.
When discovery is mediated by LLMs, ranking a page is no longer sufficient because the system is not choosing a single best document. It is assembling an answer from spans that fit its internal representation of the problem, which quietly invalidates many of the assumptions SEO and ads were built on.
You can see this shift in the kinds of services being offered now. Instead of focusing on links, keywords, or bid optimization, teams are spending time on structured content that breaks cleanly into answerable fragments, on entity relationships and schema that make concepts legible to models, and on persona driven content that anticipates how questions will be interpreted rather than how pages will rank.
Measurement is shifting as well. Instead of impressions and clicks, people are running prompt level tests, checking whether their content shows up across different models, and tracking inclusion and citation patterns rather than traffic. In many cases, strong traditional pages disappear entirely from answers while smaller, better structured sources surface.
From the outside this looks like traffic declining. Internally it feels more like a loss of observability, where you cannot tell whether you were excluded, partially used, or summarized away into latent knowledge.
Google will likely face the same issue as AI Mode expands. Generating answers is not the hard part. Defining what visibility means when the retrieval layer is no longer exposed is.
When discovery is mediated by LLMs, ranking a page is no longer sufficient because the system is not choosing a single best document. It is assembling an answer from spans that fit its internal representation of the problem, which quietly invalidates many of the assumptions SEO and ads were built on.
You can see this shift in the kinds of services being offered now. Instead of focusing on links, keywords, or bid optimization, teams are spending time on structured content that breaks cleanly into answerable fragments, on entity relationships and schema that make concepts legible to models, and on persona driven content that anticipates how questions will be interpreted rather than how pages will rank.
Measurement is shifting as well. Instead of impressions and clicks, people are running prompt level tests, checking whether their content shows up across different models, and tracking inclusion and citation patterns rather than traffic. In many cases, strong traditional pages disappear entirely from answers while smaller, better structured sources surface.
From the outside this looks like traffic declining. Internally it feels more like a loss of observability, where you cannot tell whether you were excluded, partially used, or summarized away into latent knowledge.
Google will likely face the same issue as AI Mode expands. Generating answers is not the hard part. Defining what visibility means when the retrieval layer is no longer exposed is.