How to track a brand in AI Overviews and AI Mode

Most teams check AI Overviews by hand. Someone searches a category question, sees the brand in the summary or doesn't, and reports the result. Then AI Mode arrives as a second place where Google writes an answer, and the same check gets repeated there.

Both checks describe one moment in one place. In a 19-day Zumi study of the same buyer question, the domains AI Overviews cited each day overlapped with its day-one sources by only 7%. AI Mode and AI Overviews also drew on mostly different pages.

AI Overviews and AI Mode (the two surfaces, in this guide) behave differently enough that tracking them well comes down to a few design choices. This guide covers each one, using what that study measured.

Key takeaways

• AI Overviews and AI Mode need separate baselines. In Zumi's study they shared 58 cited domains, while AI Overviews cited 129 and AI Mode cited 276.
• One check is one day. 77% of the domains AI Overviews cited appeared in a single answer and never returned.
• AI Mode cites about twice as many sources per answer, so a citation share on AI Mode is not comparable with one on AI Overviews.
• A link is not a mention. One domain was linked in 18 of 25 AI Overviews answers but mentioned in the text of only 7.
• Search Console shows impressions for both surfaces combined. It does not show the answer, the brand mention, or which surface produced it.

Where do these numbers come from?

The findings below come from one Zumi study, published as Gemini, AI Overviews, and AI Mode cite differently. It asked one buyer question every day from July 20 to August 7, 2026, in the United States and India.

One question in one category is a narrow base. The numbers describe what this study showed, and they are worth checking against a brand's own category before anything is built on them.

Should AI Overviews and AI Mode be tracked as one engine?

Google runs both surfaces, so it is tempting to report one "Google AI" number. The study argues against that.

AI Overviews cited 129 distinct domains across its answers and AI Mode cited 276, but only 58 appeared on both. A brand cited often in AI Overviews can still be missing from AI Mode, and the reverse.

The two surfaces also behave differently: AI Overviews sits above the regular results on some searches, while AI Mode is a separate conversation where each follow-up gets its own answer. They belong in two columns, each with its own baseline.

Which questions should be tracked?

A tracking program is only as good as its questions. The useful ones are worded the way a buyer would ask, across the stages of a purchase: category questions, comparison questions, and questions about a specific need.

The wording stays fixed once tracking starts. The study held its question word for word for 19 days, and the answers still changed every day. Changing the questions as well makes it impossible to tell what moved.

For AI Overviews, each question needs to actually trigger a summary. Many searches still return a plain results page. A question that never shows an AI Overview adds runs without adding data.

How often does a question need to run?

AI answers change between runs, even with identical wording. The study shows how much.

When each day's cited domains were compared with day one, the overlap was 7% on AI Overviews and 15% on AI Mode. Across the window, 77% of the domains AI Overviews cited and 72% of those AI Mode cited appeared in one answer only.

A single check, then, is closer to a sample than a reading. Each question needs repeated runs on a fixed schedule, with movement judged over weeks. A brand that appears in a third of answers is in a different position from one that appeared once on the day someone looked.

Is a cited link the same as a mention?

An AI answer can link to a page without saying the brand's name in the text, and it can mention a brand while linking somewhere else. Those are two different signals, and they belong in two different columns.

The study's most cited domain shows the gap. AI Overviews linked to it in 18 of 25 answers, but mentioned it in the answer text in only 7. AI Mode linked to it in 20 of 26 answers and mentioned it in 11.

A team counting links alone would report that page as fully present in AI Overviews. A buyer reading the summary saw the brand's name in fewer than a third of the answers. Both need tracking, per surface, using the measures described in GEO metrics beyond mention rate.

Can both surfaces share one target?

AI Mode gives longer answers with more sources. The study's figures for the two surfaces, side by side:

Both surfaces cited 58 of the same domains.

That changes what a good number looks like. A brand's share of citations will usually look smaller on AI Mode simply because each answer holds twice as many sources. Comparing the two surfaces against one target makes AI Mode look worse than it is.

Each surface needs its own starting point, taken from the first weeks of data, with change measured against it. The baseline guide covers how to set one.

What does Search Console show?

Google launched generative AI performance reports in Search Console on June 3, 2026, and extended them to all websites on August 31, 2026. The reports show impressions, pages, countries, devices, and dates for generative AI features on Search, including AI Overviews and AI Mode.

That is useful context. It shows which pages Google surfaced inside AI features, and when. It does not split AI Overviews from AI Mode, does not show the answer text, and does not say whether the brand was mentioned or a competitor was.

Search Console confirms which pages appeared. Tracked answers show what the answers actually said.

Does region matter?

Less than the surface, in this study. United States and India answers on AI Overviews and AI Mode were close to identical on every measure taken.

That held for one English-language question. Brands selling in markets with different languages or local competitors should test region early rather than assume either way.

What does this tracking not show?

Tracking AI Overviews and AI Mode shows where a brand stands in Google's answers. It does not show traffic, revenue, or why a buyer chose a competitor. Those sit in analytics and sales data, and the two should be read side by side rather than merged into one figure.

It is also not the full picture of AI search. Buyers also ask ChatGPT, Perplexity, and other engines, and each cites its own set of sources.

Zumi tracks Google AI Overviews on every plan, and AI Mode on Enterprise, with each engine reported on its own. Book a demo to see a brand's own baseline.