If your SEO dashboard still treats “sessions” as the win, 2026 is going to feel confusing fast: AI Mode and AI Overviews suppress clicks for different reasons, and they need different content and measurement plans.

If your SEO dashboard still treats “sessions” as the win, 2026 is going to feel confusing fast. AI Mode and AI Overviews suppress clicks for different reasons, and they need different content and measurement plans.

Here’s the constraint: in AI Mode, one 2026 source reports ~93% zero-click behavior, and another reports only 6–8% of sessions end with an external click. In other words, the “visit” is no longer the default outcome. It’s the exception. (Sources: [1][4])

But that’s not the whole story. AI Overviews show a different pattern: one source cites ~83% zero-click when AI Overviews are present, yet user behavior looks more like browsing and validation than decision closure—people scan, expand, reread, compare. (Sources: [1][7])

One primary move: split your search strategy into Cited (AI Mode) vs Considered (AI Overviews), then run one measurement-backed experiment per bucket.

The behavior split: AI Mode “picks,” AI Overviews “shops”

AI Mode is being positioned by Google product leadership as a flow that lets a user move from an AI Overview into a back-and-forth conversation—follow-up questions, no need to restart the search. That’s a closed loop by design. (Source: [4])

So the user’s job changes. Instead of “find a page,” it becomes “accept an answer.” The research brief cites a user-behavior study summary: in AI Mode, users take the AI short list as-is 88% of the time, and 74% choose the #1 item. That’s not browsing. That’s selection. (Source: [7])

AI Overviews behave differently. The same study summary describes a “Netflix browse” pattern, with users scrolling backward nearly 50% of the time to reread and compare options. And another report says 7 in 10 users only read the first few lines—yet 88% click “show more” even if they don’t intend to go deep. That’s a weird combo: curiosity clicks, shallow reading. (Sources: [7][2])

Put those together and the takeaway is uncomfortable: AI Overviews can influence buyers without sending you traffic, while AI Mode can pick winners without ever showing the buyer your site. Different surfaces, different psychology, same scoreboard problem.

Why this matters right now: Google is building a search that doesn’t need your page

At Google I/O 2026, Google said AI Overviews reaches 2.5B monthly active users. That’s not an “early feature.” That’s the interface. (Source: [2])

And the UI direction matters as much as the model. Reporting on Google I/O 2026 indicates Search is adding more generative elements—tables, calculators, simulations, dashboards, visuals—that users can interact with directly on the results page instead of leaving. The click isn’t being “stolen.” It’s being made unnecessary. (Source: [1])

Meanwhile, the May 2026 core update started rolling out May 21, 2026 and was expected to take about two weeks. Early coverage suggests a continued preference for first-party, expert-led content over thin or lightly edited AI content. So yes, content quality still matters. It’s just being consumed in a different place. (Sources: [1][3][4])

The experiment: build “Cited vs. Considered” pages and measure lift, not clicks

Here’s the practical framing for a CMO/VP Marketing: treat AI Mode as a shortlist engine and AI Overviews as a comparison surface. Then design content to win each job.

Step 1 — Define two page types (don’t blend them): a Cited page is written to be extracted cleanly into an AI answer; a Considered page is written to help a human validate and choose after the summary.

Cited pages should be tight: definitions, constraints, clear “when to use / when not to use,” and evidence-backed explanations that can be lifted into a summary. This aligns with the guidance in the research brief: write for summary extraction (concise answer blocks) and include value AI can’t fully replace, like firsthand experience and unique data. (Sources: [5][6])

Considered pages should assume skimming in the first lines, then reward deeper expansion: structured comparisons, decision criteria, trade-offs, and “what to check next.” That matches the AI Overviews behavior: skim, expand, reread, compare. (Sources: [7][2])

Step 2 — Make the hypothesis falsifiable: “If we publish one Cited page for each high-intent category query and one Considered page for each ‘vs/alternatives/pricing’ query, then qualified pipeline influenced by organic will hold steady (or rise) even if organic sessions fall, because more buyers will make vendor shortlists and evaluation decisions directly from AI surfaces.”

Directional, not definitive. But it’s testable.

Step 3 — Change the scorecard (or you’ll make the wrong call): the research brief is explicit that optimization is shifting from ranking and clicks toward visibility inside AI answers—citation, trust, share of voice—because users may consume summaries and never click. (Sources: [3][6])

Also, beware over-reading platform dashboards. A drop in last-click organic doesn’t prove organic stopped working. It can just mean the influence moved up-SERP. The better readout is incrementality: did pipeline change relative to a baseline or holdout?

Run it this week: one sprint, one category, one holdout

Here’s the 5-minute version you can run this week:

Success = stable or improving qualified pipeline from organic-assisted paths for the test category versus holdout. Guardrails = branded search demand doesn’t dip materially; demo-to-SQL conversion rate doesn’t degrade (quality check). Stop-loss = if assisted pipeline and demo rate both drop for the test category versus holdout at the 28-day mark, revert and rework the content structure.

Trade-off (say it out loud): this will probably reduce reported organic traffic before it improves measurement quality. That’s the point. The team is swapping a comforting metric for a useful one.

AI Mode and AI Overviews aren’t “two flavors of the same SERP.” They’re two different user jobs: pick vs. validate. The teams that win in 2026 won’t be the ones who argue about whether zero-click is “good” or “bad.” They’ll be the ones who design for the new behavior—and measure impact without needing the click as proof.