How to Show Up in Google AI Overviews and AI Mode (2026)

AI Overviews is the AI surface with the most raw traffic behind it, it sits on top of ordinary Google searches, and AI Mode extends the same machinery into a full conversational results experience. It is also the engine where GEO and classic SEO overlap most: Google's models answer from Google's index, so ranking still buys you a seat at the table. The catch is that ranking alone no longer buys the citation.

How AI Overviews chooses its sources

Two filters, in order. First, retrieval: the model pulls candidate pages overwhelmingly from what Google already ranks for the query and its close variants, page one helps, but 62% of AI citations come from beyond page one, so a top-20 presence across the cluster is the real entry ticket. Second, extraction: from those candidates it quotes the passages that read as complete, self-contained answers. Pages that rank but ramble get skipped for pages that rank slightly worse and answer cleanly.

Freshness compounds here too: 85% of AI Overview citations are from content published or updated within two years, and nearly half from the current year. An old high-ranking page is a melting asset; refresh it and it re-qualifies.

The eligibility checklist (misses here are silent)

• Snippets enabled. No blanket nosnippet or restrictive max-snippet on money pages, if Google can't quote you, AI Overviews can't cite you.
• Googlebot and Google-Extended allowed in robots.txt, and key pages server-rendered with a 200 status and a self-referencing canonical.
• Schema that matches the visible text: Article or FAQPage with an honest dateModified. Schema doesn't win the citation by itself, but mismatched or stale markup quietly disqualifies.
• Core Web Vitals in the green, AI Overviews draws from ranked pages, and vitals are part of why pages rank.

The content playbook

• One question per H2, answer first. Open every section with the 40–60 word direct answer, then the evidence. This is the passage structure AI Overviews lifts.
• Cover the query cluster, not just the head term. AI Mode especially fans out into follow-up questions; a page (or cluster) that answers the follow-ups gets cited across the whole conversation.
• Use lists and tables for anything comparative or procedural, the formats extraction favors.
• Add sourced statistics and named quotes, the same Princeton-measured levers (+41% and +28%) that work everywhere work here.
• Refresh quarterly and mean it, real content updates with a bumped dateModified, not date-stamp theater, which Google detects and discounts.

What's different from ChatGPT and Perplexity

Google's AI surfaces lean harder on your own site than the other engines do, brand-owned pages make up a much larger share of citations, which means on-page work pays off relatively more here, while ChatGPT and Perplexity lean on third-party consensus (reviews, communities, editorial). That's why a real GEO program runs both tracks at once: rank-and-structure for Google's surfaces, off-site citation building for the rest. The full split is in our GEO vs SEO guide and the Perplexity playbook.

Measuring it

Since mid-2026, Search Console reports AI Overviews and AI Mode impressions directly (Performance, Search Generative AI), which makes this the one AI surface with first-party data. Watch those impressions alongside your prompt panel, and treat rising AIO impressions on a page as the leading indicator that extraction is working. Baseline yourself first with the free AI Visibility Checker, then track weekly, our measurement guide has the full setup.