How to get cited in Google AI Overviews

Google AI Overviews are the highest-stakes surface in search, for a blunt reason: they occupy the pixels that used to belong to the #1 result, on queries that already had buying intent. When an Overview answers the question and cites three sources, the old organic results start their day below the fold. We check Overviews daily across client portfolios; here’s what the logs actually show.

How Overviews pick their citations

AI Overviews are grounded in Google’s live index, the generated text is stitched from pages Google already ranks and trusts for the query cluster. That makes them the most SEO-adjacent of the AI surfaces: your existing rankings are the qualifying round. But selection within the qualified pool follows its own logic, and this is where we see sites win or lose:

• Sub-question fit. An Overview is assembled from parts: definition from one source, costs from another, caveats from a third. Pages that cleanly own one sub-question get cited for it. Pages that gesture at everything get skipped for each part.
• Extractability. The winning passage is usually a self-contained 40–80 word block that answers without needing the surrounding text. Headings that state the question, then an immediate direct answer, keep showing up in the citation chips.
• Concrete claims. Ranges, prices, timelines. “Implant costs typically run $3,000–$4,500 per tooth” is Overview bait. Unfalsifiable marketing prose never gets picked, no matter where it ranks.

The playbook that’s worked for us

1. Find your trigger set. Check which of your priority queries currently show an Overview at all. In our client data the share varies wildly by niche, from under 20% in some local categories to most of the informational queries in health-adjacent ones. There’s no point optimizing for a box that doesn’t appear.
2. Read the assembled answer, not just the chips. Map which sub-claims the Overview makes and who supplied each. The gaps, sub-claims sourced from weak pages, generic publishers, or forums, are your entry points.
3. Restructure your page to own one part. Question as H2, direct answer in the first sentence, supporting specifics after, schema matching the content. We’ve watched this single change move a page into the chips within two index refreshes.
4. Keep the page warm. Overviews show a visible preference for recently updated sources on time-sensitive topics. The mechanics are in our freshness write-up.
5. Re-check weekly. Overview citations rotate more than any other engine’s. A monthly snapshot will randomly tell you you’ve won or lost; only the weekly trend is real. Method here.

What not to bother with

We’ve found no evidence that AI-specific meta tags, “AI optimization” plugins, or prompt-injection-style text aimed at crawlers does anything except embarrass you in your own source code. Overviews read the same page humans do. The leverage is in answer structure, claim specificity, and freshness, the same levers as the rest of the GEO playbook, applied to the engine with the most riding on it.