I've Tried Everything But AdTech Therapy

Google Won. Now What?

Jana Meron and Scott Messer of Ad Tech Therapy joined AdExchanger's Anthony Vargas for the 500th episode of AdExchanger Talks and spent most of it delivering the same message from different angles: nobody is coming to rescue publishers. The judge left Google's ad stack intact. AI Overviews are eating referral traffic. And a quiet Google Ad Manager accounting change is about to lop up to 25% off reported impression volume. The through-line is that the "wait for the breakup" option is dead, and the clock is running.

The remedy changed nothing

Messer's read on the antitrust outcome is blunt: even with behavioral remedies, Google "still controls the demand ultimately." Log-file access gives publishers visibility, not dollars. Independent SSPs that built their pitch around a forced divestiture now have to beat an intact incumbent on merit. Watch which ones quietly stop mentioning the remedy in their decks.

The Amazon FTC suit is a separate matter and worth less attention from open-web publishers than it's getting. The allegation there is ghost bids inside Amazon's own retail auction. Different mechanism, different harm. It barely touches open-web supply.

The impression-count change lands Tuesday morning

Google Ad Manager is switching from counting an ad when it downloads to counting it when it begins to render. Messer has seen the gap hit 25% on real publishers. Traffic doesn't change. Reported impressions do. Every CPM guarantee, every year-over-year yield comparison, and every make-good calculation now has a discontinuity in it. Get ahead of the number before your finance team discovers it on a dashboard and demands an explanation you haven't prepared.

AI is the structural problem; agentic pipes are not the fix

No publisher on Meron and Messer's roster reported a traffic gain from generative search. None. AI Overviews tell users the answer and kill the click. That is the signal underneath everything else. The agentic AI enthusiasm, MCP servers, Meta agents buying autonomously, is real plumbing. But Messer made the gap plain with a live client example: the SSP showed up and asked the publisher to bring the demand. Connectivity without liquidity is not a market.

The only structurally sound bet

Media quality tools like OpenSincera (The Trade Desk's supply-chain transparency product) show publishers where they stand. The problem is that improving the score means cutting reseller SSPs and reducing ad density, which means a revenue hit this quarter for better demand routing next year. Most publishers can't afford the dip. Meron's newsletter-first cohort, publishers building direct audience before building a website, is the one model that sidesteps the problem entirely. Owned distribution is the only line item AI can't intermediate.

Our call: Before the May 2027 upfront season, at least two mid-to-large ad-supported publishers will publicly attribute a year-over-year impression decline to Google's download-to-render counting change. The gap is large enough (25% in documented cases) to trip year-over-year comparisons that investors and parent companies will notice. "Google changed how it counts" is the only true explanation available, and public-company reporting makes quiet absorption unlikely.

Google beat the ad-tech breakup, and the two people who advise publishers for a living called it a "big nothing." That should tell you how little the antitrust fight was ever going to change your P&L. On the 500th AdExchanger Talks, Jana Meron and Scott Messer of Ad Tech Therapy sat with Anthony Vargas and walked through the real menu of publisher problems: a judge who left Google intact, AI answers eating referral traffic, and a quiet accounting change in Google Ad Manager that will lop up to a quarter off your reported impressions. The through-line is that nobody is coming to help publishers. They have to fix the machinery themselves.

Here's how the council reads it.

What's actually being decided: Two things, and only one of them is a decision. The Google remedy and the Amazon suit are events you react to. The real choice for operators is where to spend scarce remediation effort now that the "wait for the breakup" option is dead. That choice is easy to undo in principle, but the traffic erosion sets the clock, so delay has a cost.

---

The Market Analyst. The market spent two years pricing in a Google divestiture as a growth catalyst for independent SSPs and ad servers. That trade is over. Scott Messer's line matters: even with behavioral remedies, Google "still controls the demand ultimately." Log-file access gives publishers visibility, not dollars. So the independents that pitched "when Google is forced apart, we win" now have to win on merit against an intact incumbent. For anyone informed but not in the weeds: the court took the plumbing away as a legal issue and left the money exactly where it was. Watch which independent SSPs quietly stop citing the remedy in their decks by the next earnings cycle.

The Skeptic. Two claims here need a harder look before anyone acts on them. First, Meron framed the Amazon FTC suit as "the same kind of scheme Google was running." It isn't. Google's case was about owning the whole stack. Amazon's is alleged ghost bids inside its own retail auction. Different harm, and it barely touches open-web publishers. Second, the Google Publisher Data Co-op toggle. Messer says publishers get no direct payment, only a vague CPM promise. Probably right, but the actual co-op terms weren't cited, and his consultancy competes with Google's built-in tools. Verify against current GAM documentation before you flip or refuse the toggle on his say-so.

The Operator. The impression-counting change is the one that lands on your desk Tuesday morning. Google moves from counting an ad when it downloads to counting it when it begins to render. Messer has seen the gap hit 25% on some publishers. Nothing about your traffic changes, but your reported impression volume drops, and you get to explain that to a management team that reads dashboards, not methodology notes. Every CPM guarantee, every year-over-year yield comparison, every make-good calculation now has a discontinuity in it. Get ahead of the number before finance discovers it. And when an SSP pitches you "agentic pipes," remember Messer's live client call: the SSP asked the publisher to bring the demand. That's an empty pipe.

The Customer / End User. Two customers here, and they pull apart. Advertisers are the ones harmed in the Amazon case, overpaying by some contested figure Messer himself couldn't pin down ($8B or $20B, he wasn't sure, so treat neither as real). The buy-side reading in the AdExchanger piece Ken flagged is that buyers are exhausted from tracking auction mechanics they can't see. The publisher customer, meanwhile, is being told by AI Overviews that they don't need to click through at all. No publisher on Meron and Messer's roster reported a traffic gain from generative search. None. That is the signal underneath everything else.

The CFO. Media quality tools like OpenSincera from The Trade Desk give you an honest mirror. The problem is the mirror shows a bill. Improving your quality score means cutting reseller SSPs and reducing ad density, which means taking a revenue hit this quarter to earn better demand routing next year. Most publishers can't afford the dip, so they stay stuck. Same math on the impression change: the volume drop is optical, but if your deals are priced on impression counts rather than outcomes, the renegotiation is real cash. The one structurally sound bet in the whole episode is Meron's newsletter-first cohort, building direct audience before building a website, because owned distribution is the only line item AI can't intermediate.

---

Where the council splits:

The biggest disagreement is whether Google walking away from open-web display is good or bad for publishers. Messer floated that Google "could start to ignore some of this business and let everybody else fight over it." The Market Analyst reads that as opportunity for independents. The Operator reads it as the demand rug getting pulled while the plumbing stays Google's. Both can be true, and which one you're living depends on how much of your revenue routes through Google demand today.

The second split: is agentic AI early or empty? Messer says there's not "nearly as much money inside Agentic as anyone thinks." The MCP-server enthusiasm (Meta agents building audiences and launching campaigns autonomously) says the plumbing is real. The Skeptic's point settles it for now: real pipes, no demand liquidity. Connectivity is not a market.

What it hinges on: Whether publisher revenue is a demand problem or a plumbing problem. The whole episode argues it's demand, which is why the remedy changed nothing and why owned audience is the only durable answer. If that's right, every hour spent on toggles and quality scores is triage, and every hour spent on direct distribution is the actual business.

The council leans hard toward demand. Verify the co-op terms and the impression gap on your own inventory before you renegotiate anything, but don't wait for structural relief. It isn't coming.

---

Prediction: Before the 2027 upfront/NewFront selling season (roughly May 2027), at least two mid-to-large ad-supported publishers will publicly attribute a year-over-year drop in reported ad impressions to Google's begin-to-render counting change rather than to lost traffic.

Confidence: Medium — the change is confirmed and the 25% gap is documented, but timing depends on GAM's rollout schedule.

Why: Google is switching Google Ad Manager from counting an ad when it downloads to counting it when it starts to render, and Scott Messer has seen that gap reach 25% on real publishers. When reported impression volume falls that much with no change in actual traffic, finance teams and public-company reporting will demand an explanation, and "Google changed how it counts" is the only true one available. Publishers that report ad metrics to investors or parent companies can't quietly absorb a double-digit optical decline, so at least a couple will name the methodology change on the record to protect their numbers. The opposite outcome, everyone absorbing it silently, is unlikely precisely because the drop is large enough to trip year-over-year comparisons that others will notice first.

Revisit by 2027-05-15: We're right if at least two publishers (in earnings commentary, trade press, or IAB/MRC forums) blame an impression-volume decline on the download-to-render change. We're wrong if the change rolls out with no publisher publicly tying it to a reported impression drop.