Refacto

Industry story

Google routes AI Overview links into AI Mode, bypassing publisher sites

ai-in-adtech big-tech programmatic publisher-economics walled-gardens

Google just turned AI Overviews into a retention mechanism. The links it's now appending to those AI-generated summaries don't send users to publisher sites; they push users deeper into AI Mode, Google's fully AI-driven search layer, framed as helpful follow-up questions. Every outbound click to a publisher was a cost to Google with no corresponding revenue. Keeping the user inside converts that giveaway into a second query, a third, and the ad revenue against all of them. Publishers running open-web programmatic should expect this in the GA4 logs before anyone names it officially: Google organic sliding, the programmatic impression pool shrinking, and no content fix that closes the gap.

Full analysis

Google is now dropping links at the bottom of its AI Overviews (the AI-written summaries that sit above the blue links) that push you deeper into AI Mode, its fully AI-driven search experience, instead of out to a publisher's website. Framed as helpful follow-up questions. The effect is the same either way: the user stays inside Google, and the click that used to land on someone's page never leaves the building.

How hard is this to undo? For Google, trivially easy. It's a UI change. For publishers, the traffic loss is hard to undo, because once a reader learns they can get the whole answer plus the follow-up without leaving, they stop clicking through for good. Habits don't reverse.

What's actually being decided: Nothing, by publishers. This isn't a decision for the ad-tech reader to make. It's a shift to react to: Google is converting search from a place that hands out traffic into a place that keeps it.

What sets the deadline: AI Mode adoption. This matters when a meaningful share of queries default into it. Not yet, but the direction is one-way.


The Market Analyst. The story here isn't Google. It's who depends on the traffic Google is now keeping. Open-web programmatic ad-tech (the pipes that sell display and video on publisher sites) lives on pageviews. Fewer referred sessions, fewer impressions, lower revenue for Magnite and PubMatic, whose inventory is largely publisher remnant. The market has already marked down open-web ad-tech multiples on the AI Overview fear, but this routing change accelerates the timeline rather than just repeating the thesis. The winners sit inside owned traffic: Amazon Ads, retail media networks, anyone who doesn't rent their audience from Google. In plain terms: money follows eyeballs, and eyeballs are moving inside walls.

The Skeptic. Steelman the calm case. AI Mode is opt-in and nascent. A follow-up prompt bolted to the bottom of a unit most users scroll past isn't force-routing anyone. Publishers were already bleeding referral traffic before this link existed, so the marginal hit from one more prompt is small this quarter. And Google is under active antitrust scrutiny in the US and EU for exactly this kind of self-preferencing. Aggressively starving the open web inside a product a court is watching creates fresh legal exposure Google's lawyers understand better than anyone. The loudest alarms come from SEO-dependent publishers whose model was already broken. They frame every Google tweak as the apocalypse because it's always been their apocalypse.

The Operator. Publisher ad ops will see this in the logs before anyone names it. GA4 shows Google organic sliding while direct holds flat, which hides the actual choke point. SEO managers benchmarking on Search Console click-through will file the Q3 drop as algorithm volatility and burn two quarters chasing content fixes for a problem that isn't a content problem. The real damage is programmatic yield: fewer pageviews, fewer impressions, and the CPM floor tricks that lean on volume don't rescue a site that's lost a fifth of its Google sessions. Display and open-web video take the first hit. Then newsletter and subscription conversion take the second, because the top-of-funnel discovery that fed them is drying up.

The CFO. Run the money and the incentive is obvious. Every outbound click to a publisher was a cost to Google, not a revenue line. Google earned on the search ad, then handed the user away for free. Keeping that user inside AI Mode means the next query, and the ad against it, also belongs to Google. So the routing change doesn't cost Google traffic revenue, because it never made money on the exit. It converts a giveaway into retained, monetizable engagement. For the publisher CFO, the mirror image: a revenue stream that was already thinning now has a structural leak, and the backup plan (direct and subscription) needs top-of-funnel that Google is quietly closing.


Where the council splits.

The Skeptic and the Market Analyst disagree on timing, and it's a real disagreement, not a hedge. The Skeptic says AI Mode is too small to matter this quarter and antitrust makes Google cautious. The Analyst says the direction is one-way and Q3 comps will be ugly regardless of how nascent the surface is. Both can be right: small now, structural over four quarters.

The second split is Skeptic versus CFO on whether Google actually wants this to go all the way. The CFO's math says Google has every incentive to keep users inside, because outbound clicks were pure cost. The Skeptic says a live antitrust case caps how far Google will push. That tension is the whole story: the money says press the loop shut, the courtroom says go slow.


What this hinges on. Two things. One, how fast AI Mode becomes the default surface for real query volume rather than an opt-in curiosity. Two, whether antitrust remedies land hard enough to force Google to keep sending traffic out. If AI Mode stays small and remedies bite, publishers get a reprieve. If AI Mode scales and remedies stay theoretical, the open-web ad-tech TAM contracts in steps, each time Google tightens the loop.

The council leans one way. The CFO math is the strongest piece on the table: outbound clicks cost Google money and earned it nothing, so the incentive to contain the user is total and permanent. Antitrust slows the pace, it doesn't reverse the incentive. Publishers who still treat Google organic as a growth channel are planning around a baseline that's already gone.

What to de-risk: stop attributing the Q3 referral drop to algorithm noise, separate Google-referred sessions in reporting so the structural leak is visible, and move budget and product effort toward owned distribution (email, app, direct) before the top-of-funnel dries out further.


Prediction: Google's referral traffic to third-party publishers, as measured by Similarweb's aggregate publisher-traffic data, will show a year-over-year decline for the fourth quarter of 2026 when that data publishes in January 2027.

Confidence: Medium. The incentive is clear, but the size of the near-term move is uncertain.

Why: Google earns nothing on an outbound click to a publisher; that click was always a cost it absorbed to keep search useful. AI Overviews already suppress click-through by answering the query on-page, and now the follow-up links route into AI Mode instead of out to sites, so each new AI surface removes another reason to leave Google. The mechanism is one-directional: once a user gets the answer plus the next question without clicking through, that referral does not come back. The opposite outcome, referral traffic rising, would require Google to reverse a change that costs it nothing and earns it retained engagement, which runs against its own economics.

Revisit by 2027-01-31: We're right if Similarweb's Q4 2026 publisher-referral data shows Google-sourced traffic down versus Q4 2025. We're wrong if it's flat or higher.

Comments