The AI Overviews playbook: what we changed in Q1 2026
Five concrete changes we shipped across client sites this quarter, and the citation-rate movement we measured. Two worked, one was neutral, two were duds.
I'd rather read a quarterly post-mortem than another speculation post. This is ours for Q1 2026 — five changes we made to client content with the explicit goal of moving AI Overviews citation rate, what we measured, and where we were wrong.
Change 1: lede rewrites for answer-shape (worked)
Hypothesis: pages whose first paragraph is a direct, declarative answer to the head query get cited more often than pages that open with context. We rewrote the lede on 84 pages across three client sites — same factual content, just front-loaded.
Result: median citation rate across those pages moved from 12% to 21% over six weeks. The effect was strongest on definitional queries ('what is X', 'how does Y work') and weakest on comparison queries.
Change 2: original data callouts (worked)
Hypothesis: pages with specific original numbers — survey results, internal benchmarks, dated observations — would get cited more often than synthesis posts. We added an 'In our data' callout to 41 pages where we had genuine internal data.
Result: a smaller sample, but a clearer effect. Citation rate on those pages went from a median of 9% to 26%. We were not measuring that those pages would also start to be cited for queries we hadn't targeted — they did. Original data appears to expand a page's relevant query surface, not just deepen its existing relevance.
Change 3: FAQ schema expansion (neutral)
We added detailed FAQ schema to 60 pages, including questions sourced from the People Also Ask boxes for each page's target query. Pre-test citation rate: 14% median. Post-test: 15%. Within the noise floor of our sample sizes.
I'd been recommending FAQ schema as a possible AI Overviews lever for a year. The data says it isn't. We've stopped including FAQ schema as a default deliverable.
Change 4: aggressive internal linking to target pages (dud)
We added inbound internal links from an average of 14 new pages to each of 22 target pages. The hypothesis was that AI Overviews would treat strong internal linking as a relevance signal, similar to classic on-site link-equity behavior. Citation rate moved 1 percentage point. We're calling that zero.
Change 5: byline pages with structured author entities (dud, but informative)
We built proper author pages — full bio, links to past work, organizational schema — for the six bylined authors on one client site. Citation rate for their pages didn't move in 90 days. But we also saw qualitative changes: when those authors were cited (rare), the citation occasionally surfaced the bio link. It's plausible the effect needs longer, or needs the author to also appear elsewhere on the web. Either way, the short-term ROI isn't there.
What we are doing in Q2
Doubling down on lede rewrites and original data, treating them as the new minimum bar for any page we want to be cited. Continuing the author-entity work but on a slower clock — six months minimum before declaring it dead. Stopping FAQ schema and aggressive internal linking as AI-targeted tactics. They have their uses elsewhere; they aren't this lever.