The answer box got bigger. Be inside it.

Observably, from pages that already rank reasonably for the query or closely related ones (retrieval leans on the existing index), then favouring content that answers the specific sub-question directly, in self-contained passages, with corroborating facts across sources and clear structure. Citations skew toward pages that state things plainly and specifically: hedged, padded or keyword-stuffed content is passed over even when it ranks, because there is nothing clean to lift.

They redistribute it, and the effect varies enormously by query type. Informational queries with simple answers lose clicks: the answer appears and the user leaves satisfied. Commercial and consideration queries behave differently, because AI Overviews frequently prompt further research, and industry studies through 2026 report that a large majority of users who encounter an AI Overview still click through to at least one cited source. The strategic response is to be that cited source and to shift content investment toward queries where a summary cannot substitute for the visit.

Four, applied per page: a direct answer within the opening lines, written to stand alone if extracted; specifics over generalities (numbers, ranges, named conditions, dated facts); headings that mirror the sub-questions the query implies; and structured data plus honest dates so the machine can verify what it is quoting. These also improve human comprehension, which is the reliable test of whether an AI-optimisation tactic is legitimate.

Adjust rather than panic. Deprioritise thin informational content whose only value was a click that AI now absorbs, invest in content whose value requires the visit (tools, calculators, deep guidance, commercial decisions, local action), and treat AI Overview citations as brand impressions with measurable downstream effect on branded search. Businesses whose entire content model was quick-answer articles do need a genuine rethink, and we say so.

Partially, and honestly: Search Console does not break out AI Overview impressions separately, so measurement combines prompt-sampling (running your money queries on schedule and logging whether an AIO appears and who is cited), CTR analysis on queries where AIO presence is known, and branded-search movement. Anyone selling a precise AI Overview rank tracker is selling estimates with confident styling.

Largely yes, and it transfers well: the structural habits that won featured snippets (direct answers, clean definitions, tables and lists where appropriate, question-shaped headings) are the same habits AI Overviews reward. The difference is that snippets rewarded exactly one winner while Overviews cite several sources, which makes the work worth doing for a wider set of pages than before.

AI Overview readiness audits from £700 including prompt-sampled baselines for your priority queries. Implementation sits inside content and technical retainers from £950/month rather than as a separate product, because the changes are page-level content and structure work rather than a distinct discipline.