Both surfaces cut clicks to the open web, and the evidence suggests AI Mode cuts more. In a 2026 randomized experiment, AI Overviews reduced outbound organic clicks by about 39.8% yet left user satisfaction unchanged. The same study's exploratory AI Mode arm left roughly 0.36 external clicks per search against 0.53 for control users, and lowered satisfaction by 1.14 points. The difference is structural: an AI Overview sits atop a normal results page, so links survive below it, while AI Mode replaces the page with a conversation.
This compares the two Google surfaces on click and traffic impact. For the separate question of which sources each one cites, see do AI Mode and AI Overviews cite the same sources. The click figures come from Agarwal and Sen's 2026 SSRN working paper (not yet peer-reviewed, US desktop Chrome), covered in full in do AI Overviews actually cause click loss; the AI Mode arm there was exploratory, so read its numbers as directional.
What is the difference between AI Overviews and AI Mode?
They are two different answer surfaces from the same company. Naming them precisely matters, because they behave differently for publishers.
- AI Overview is a generative summary that appears above the classic list of blue links on a normal Google results page. The organic links still exist below it, so a searcher can scroll past the answer to them.
- AI Mode is a separate, fully conversational interface. It answers in a chat-style thread, surfaces fewer outbound links, and supports follow-up questions rather than a fixed list of results.
The practical consequence is that an AI Overview leaves more of the old results page intact, while AI Mode is closer to a self-contained answer that a searcher rarely leaves.
Which surface takes more clicks?
On the available causal evidence, AI Mode leaves fewer clicks to the open web, though the comparison is not like-for-like. The table below sets the measured figures side by side, each with its own baseline.
| Measure | AI Overviews | AI Mode |
|---|---|---|
| Outbound clicks (shown vs counterfactual) | 0.37 per search shown vs 0.62 hidden (~39.8% cut) | ~0.36 external clicks per search vs 0.53 for control |
| Effect on user satisfaction | No meaningful change | Fell by 1.14 points |
| Surface layout | Summary above surviving organic links | Full conversational takeover, fewer links |
| Evidence strength | Primary randomized arm | Exploratory arm (forced redirect) |
Source: Agarwal and Sen, 2026 (SSRN working paper). The two click figures use different metrics and subsamples, so the honest read is directional: AI Mode both removed more outbound clicks and lowered satisfaction, whereas AI Overviews removed clicks without changing how satisfied users felt. Do not subtract one number from the other as if they were the same measurement.
How do they compare on citations, not just clicks?
Differently again, which is why you track both. Click impact tells you how much traffic a surface removes; citation sourcing tells you which pages it names as sources. On sourcing, two 2026 studies found the two surfaces share only about 59% of their top cited sources, and AI Mode drew far less from the classic organic top rankings (roughly 19% of AI Mode citations from the organic top-20 versus about 56% for AI Overviews). The full breakdown is in do AI Mode and AI Overviews cite the same sources.
AI Overviews and AI Mode are two zero-click surfaces with different physics: the overview thins the clicks on a page you still recognise, while AI Mode replaces the page and cites a different set of sources. Winning one does not mean winning the other.
What does this mean for your visibility strategy?
Plan for both as zero-click surfaces, and shift the goal from the click to the citation. Concretely:
- Expect a structural click drop on both. Neither surface lets you opt out of appearing, and both remove outbound clicks on informational queries, so budget for it rather than treating it as a temporary dip.
- Compete to be the cited source on each surface separately. Because they cite different pages, verify your presence on AI Overviews and AI Mode independently, not as one score.
- Measure citations alongside traffic. If sessions are your only metric, both surfaces read as pure loss even while your brand is shown and recommended, the shift laid out in from clicks to citations.
The two surfaces differ in how they take the click and in which sources they cite, so a single traffic number can hide whether you are still winning the answer. Watching where you are cited, where you are merely shown, and where AI still sends visitors, across both surfaces and over time, is exactly what Buffy Intel is built to do. Questions: [email protected].