If organic sessions are flat (or down) while impressions look fine, AI Overviews are a likely culprit. The fix isn’t “publish more.” It’s treating Google’s results page like a decision surface you have to measure and win.

If organic sessions are slipping while Search Console impressions look healthy, the problem might not be rankings. It might be the page you never owned in the first place: Google’s.

In a randomized field experiment covered by Search Engine Journal, AI Overviews were associated with a 38% reduction in organic clicks on queries where they appeared, and zero-click behavior rose from 54% to 72%. That’s not a rounding error. That’s a new funnel shape.

Now for the part that should make every CMO pause: in the same coverage, removing AI Overviews didn’t materially improve satisfaction, perceived information quality, or ease of finding information. So the user doesn’t feel worse. They just leave your site out of the journey more often.

That’s the constraint. Here’s the outcome to aim for in 2026: keep qualified pipeline stable even as SEO clicks get rationed.

The nut graf: Google is becoming the “landing page”

AI Overviews expanded to all U.S. users in May 2024, shifting from an experiment (SGE) to a default layer in Search. The practical result is simple: more evaluation happens on the SERP, before anyone hits a vendor site.

A study covered by Search Engine Journal analyzed 846,000 U.S.-based Google search sessions (from February and March 2026) and found that when an AI Overview is present, users show more on-SERP engagement: more time comparing, more scrolling back up, and more “still cursor” behavior consistent with reading instead of scanning.

Google has also said—per Search Engine Journal’s reporting—that AI Overviews drive more than 10% additional queries for the types of searches where they appear. More searching. Fewer clicks out. That’s the trade.

What 846,000 searches say about attention (and why intent segmentation breaks)

The ClickStream analysis (as summarized in the provided source content) used cursor tracking as a proxy for attention, sampled at one-second intervals, across informational, local, navigational, transactional, and video searches. The consistent story: AI Overviews slow people down.

Look at “active at 21 seconds” as one rough read on how long users stick around on the SERP. Without AI Overviews, navigational searches (often brand-name queries) had only 12.0% of users still active at 21 seconds. With AI Overviews, that jumped to 45.8%. Informational went from 21.6% to 45.4%. Transactional from 24.9% to 47.4%.

That’s the pattern interrupt: even when someone types a brand name, AI Overviews can pull them into comparison mode. The SERP becomes a mini “category page,” not a direct route to a homepage.

And there’s another twist the Search Engine Journal coverage called out: when an AI Overview is present, differences in behavior by intent type largely disappear. In plain English, the AI Overview becomes the dominant driver of behavior—not whether the query was informational, navigational, or commercial.

If the team’s SEO reporting still leans on intent buckets to predict outcomes, that model is going to miss. Badly.

The one move: build a SERP holdout to measure incrementality

If clicks are falling and the user experience isn’t, the only sane question is: what’s the incremental value of organic visibility now?

Here’s the 5-minute version you can run this week: create a measurement loop that treats AI Overview presence as a “treatment,” then reads impact downstream in pipeline—not in last-click sessions.

The hypothesis (make it falsifiable): If we segment SEO performance by AI Overview presence and run a geo or keyword holdout, then qualified pipeline attributed directionally to organic + brand will shift (even if clicks don’t) because the SERP is absorbing early-stage evaluation and redistributing demand across later touchpoints.

But the data tells a different story than most dashboards: Search Console can show rising impressions while GA shows shrinking organic sessions, and both can be true. That’s not “tracking issues.” That’s user behavior moving up-funnel onto Google.

Run it this week: setup / launch / readout / next test

Success = +5–15% lift in qualified pipeline in treatment vs control (directional target; calibrate to your volume). Guardrails = no increase in sales cycle length and no drop in SQL-to-close rate. Stop-loss = if branded demand and assisted opportunities fall for two consecutive weeks in treatment, revert changes and re-scope the keyword set.

Trade-off (say it out loud): This will reduce volume before it improves quality in some accounts. When the SERP becomes the evaluation layer, fewer low-intent clicks is normal. The question is whether the remaining demand converts better.

When this is wrong: If your motion depends on high-volume, low-consideration traffic (ads-supported, self-serve, or massive top-of-funnel), losing clicks can be existential. In that case, the “SERP holdout” still matters—but the mitigation may be shifting budget into channels where you control the landing experience.

The kicker: the click was never the product

The 846,000-search dataset and the experiment results Search Engine Journal covered point to the same uncomfortable truth: AI Overviews don’t just change where traffic goes. They change what the SERP is. Reading behavior increases (cursor still 44% vs. 29% without AIO in the ClickStream summary). Back-scrolling rises (51% to 59%). People compare. Then they decide.

So the job in 2026 isn’t to “get the click back.” It’s to prove, with a holdout and real pipeline math, whether visibility on Google still creates incrementality when Google keeps more of the journey for itself.