A search that used to send a click now often ends inside the results page itself. An AI Overview reads several sources, synthesises an answer, and shows it above the traditional results — and for a large share of informational queries, that’s where the searcher’s journey ends. Ranking first no longer guarantees a visit. That doesn’t make keyword strategy obsolete; it makes some keywords far less valuable than they used to be, and others more valuable than ever.
AI Overviews cannibalise the queries they used to send you
The queries most likely to trigger an AI Overview are exactly the broad, informational ones classic keyword research has always prioritised — “what is,” “how does,” “best way to.” Ranking well for those is still worth something for authority and for the AI Overview’s own citations, but the click that used to follow a top ranking increasingly doesn’t arrive, because the answer is already on the page. Treating every keyword with search volume as equally worth chasing stopped making sense the moment volume and clicks decoupled.
Sort keywords by “answerable” vs. “decision”
The keywords still reliably sending traffic are the ones an AI Overview can’t fully answer on its own — anything comparative, transactional, or specific to a decision the searcher still has to make. “Best CRM for a 10-person team,” a specific product comparison, a query with local or pricing intent — these require judgement, current information, or a transaction the AI can’t complete inline, so the searcher still clicks through. The practical shift is weighting keyword research toward decision-stage and long-tail queries, and treating high-volume broad informational terms as a citation target rather than a traffic target.
Being the cited source matters more than ranking first
When the AI Overview is where the answer lives, the win isn’t rank one — it’s being one of the sources it cites. That’s a different discipline from classic on-page optimisation, though it shares its foundations: clear structure, genuine expertise, content a crawler can parse and trust. It’s the same ground Generative Engine Optimization covers, and it means content built to be extracted and quoted — answering the question plainly, early, with structure a model can lift cleanly — earns visibility even on queries where a click was never coming anyway.
The keyword list still exists — the scorecard changed
None of this means abandoning keyword research; it means reading the results differently. Search volume alone no longer predicts value — a term’s “clickability” once an AI Overview appears matters just as much, and that varies keyword by keyword, not just industry by industry. Pairing classic rank tracking with visibility inside AI answers gives a fuller picture than either alone, and it’s exactly what we’ve argued for when it comes to earning citations rather than assuming a ranking automatically converts to a visit. Run well, SEO and AI-search visibility inform the same content strategy — they just need to be measured as two different outcomes now, not one.