Search stopped sending readers. Now it answers instead
For two decades the deal was simple: you wrote something useful, search sent you readers. That arrangement is quietly ending.
Google's AI Overviews now appear in roughly 48% of searches and reach more than two billion people monthly, making AI-generated answers a primary discovery channel — and an answer displayed on the results page is an answer nobody clicks through to read.
The consequence is uncomfortable for anyone who publishes. Being ranked first no longer guarantees being read. Being cited inside the answer is what carries visibility now, and citation follows different rules than ranking did: clear structure, verifiable claims, distinct information that cannot be assembled from ten other pages.
What this changes in practice:
-
Generic explainers lose their value fastest. If a model can synthesise your article from three other sources, it will, and no one arrives.
-
Original reporting, primary data and firsthand accounts hold up. They cannot be paraphrased from elsewhere because they do not exist elsewhere.
-
Direct channels matter more than they did. Newsletters, apps and returning readers are traffic nobody can intermediate away.
-
Structure is now machine-facing. Clear headings, explicit claims and dates make a page easier to cite, not just easier to skim.
None of this means search is dead. It means the traffic that came free for twenty years was a subsidy, and the subsidy is being withdrawn.