What exactly did Google merge?
At Google I/O 2026, Google announced the unification of AI Overviews — the AI panel that appears above classic results — and AI Mode, the conversational tab launched earlier. Both now run on Gemini 3, which has been the default model since January 2026, and both share the retrieval stack: query fan-out across sub-queries, grounding against cross-checked sources, generation, and citation. Alphabet’s Q2 2026 report put AI Overviews usage above 2.5 billion monthly users, a scale that made maintaining two diverging systems pointless.
What still differs between the two surfaces?
| Aspect | AI Overviews | AI Mode |
|---|---|---|
| Context | Inline in classic search results | Standalone conversational tab |
| Session shape | Single query, single answer | Multi-turn dialogue with memory |
| Fan-out depth | Fixed per query | Compounds across follow-up turns |
| Typical intent | Quick informational resolution | Exploratory research and comparison |
| Citation display | Link cards beside the panel | Inline links plus cards across turns |
Why does the merger simplify GEO?
Because the pipeline is one. The passage that wins grounding in an AI Overview is the same passage retrieved in an AI Mode session; the facts that fail consistency checks fail in both. Before the merger, teams debated separate “AI Mode optimization”; after it, the correct posture is one content core — answer-first sections, dated data, comparison tables, consistent facts — measured across both surfaces with the same question set. The full mechanics are in Google AI Mode and Google AI Overviews.
What does multi-turn change in practice?
AI Mode’s follow-ups extend the fan-out tree in real time: a session that starts “best CRM for agencies” continues into pricing, migration effort and integrations. Pages covering the whole tree get cited across turns; pages answering only the head question appear once, if at all. This makes query fan-out mapping and semantic coverage mapping the highest-leverage disciplines for the merged system — Semantic Coverage is, after all, one of the four measurement dimensions.
What should you actually do differently?
Mostly, nothing new — but do the existing things completely. Cover sub-questions, not just head queries. Keep facts consistent, because multi-turn grounding re-checks them repeatedly. Measure with prompts that include follow-up turns, not only single questions. And watch the zero-click math: with roughly 68% of sessions ending without a click (Ahrefs, February 2026), AI Mode’s longer sessions concentrate even more influence inside the answer. Related analysis: zero-click 2026.
A final practical note on evidence: when you measure the merged system, log which interface served each answer. AI Overviews and AI Mode share a pipeline but not a session shape, and separating the two in your screenshot archive lets you see whether multi-turn coverage work is paying off in AI Mode specifically. That split takes one extra column in the measurement sheet and saves a lot of arguing about what moved.
There is also a measurement nuance worth flagging: AI Overviews still appears on queries where AI Mode is never opened, so the panel remains the first AI impression for most users. Treat Overviews coverage as your reach metric within Google and AI Mode citation across turns as your depth metric. Both roll up into the same four dimensions, but separating them in reporting shows whether you are winning the first glance, the research session, or both.