The Remedies Episode

Full analysis

The DOJ just unsealed its 108-page playbook for breaking Google's grip on ad-tech, and Adam Heimlich and Gareth Glaser spent an episode arguing over whether it actually does anything. The structural fix sounds clean: force Google's exchange (AdX) and its buy-side network (AdWords) to bid into Prebid, the open-source header-bidding plumbing publishers already run, on the same terms as everyone else. Do that, and Gareth Glaser claims up to 80% of Google Ad Manager (GAM) users could walk, because they only ran GAM to unlock Google demand in the first place. Then there's the catch Scott Messer found. So what should an operator actually do about it?

What's being decided: whether the remedy opens the sell-side ad server market to real competition, or whether one sentence in the document lets Google keep the status quo while looking compliant. This is hard to undo once it's set. A final remedy is a court order, not a product roadmap you can revise next quarter.

The Market Analyst

Terry Kawaja's "window in a stuffy room" line is the money quote here, and it points at capital, not code. Adam Heimlich's evidence is that Beeswax was the only new demand-side platform (the software advertisers buy through) built during a multi-year stretch when DSPs were otherwise booming. That's the chilling effect, quantified. If the remedy sticks, the bet isn't that publishers save money on GAM. It's whether investors start funding sell-side ad servers again, because there's finally a market to win. Watch where the money moves, because that tells you whether the Street believes the remedy has teeth. In plain terms: a real structural remedy creates new companies; a paper remedy creates nothing.

The Skeptic

Scott Messer found the whole ballgame in pages 70 to 85. AdWords can't favor AdX because they share an owner, but it can favor AdX whenever Google says AdX delivers better ROI, better fraud protection, or better privacy. Google keeps its first-party data signals to itself. So AdWords "organically" bids more into AdX than into Prebid, and no publisher in their right mind leaves GAM to earn less. That's not a loophole Google has to sneak through. It's written in. Gareth Glaser called it "a nasty however" and he's right. The 80% exodus assumes publishers act against their own revenue, which they never do. In plain terms: the remedy bans the crime and then licenses the same behavior under a different name.

The Operator

The 600-millisecond latency claim is the part I'd stress-test before I built anything on it. Adam Heimlich states it as fact, but he'd just finished describing how messy publisher-side legacy setups are. You can't have both a universal 600ms number and infinite implementation variety. Even so, the direction is real. GAM's JavaScript wrapper does add drag, and stripping it would help viewability. But here's what breaks Tuesday morning: the 80% who supposedly don't need an ad server still need trafficking, reporting, and reconciliation. "Drop GAM" is not a button. It's a migration project with no obvious destination server, because the competitors the remedy is supposed to create don't exist yet. In plain terms: you can't leave the house before anyone's built the new one.

The Customer / End User

Put yourself in the publisher's seat, because that's who this remedy claims to help. What they actually want is more money per page and fewer middlemen taking a cut. The remedy offers a maybe on the first and nothing guaranteed on the second. If AdWords bids less outside GAM, the publisher's choice is simple math: stay and earn more, or leave on principle and earn less. Principle loses every time in this business. And the OpenAI thread is a useful reality check on the same audience. Capital One and Expedia poured direct-response budget into ChatGPT and got "diabolical" results, because people read and work in ChatGPT, they don't scroll and impulse-click. In plain terms: buyers follow returns, not narratives, on both sides of this story.

The CFO

GAM costs money and the 80% who have no direct-sold campaigns are paying for software they were forced to use. That's the clean CFO case for leaving. But run the full number. The GAM fee is small next to the revenue AdWords delivers through it. If leaving GAM means AdWords bids meaningfully lower into your Prebid setup, you've saved a rounding error and lost real yield. The payback math only works if a rival sell-side ad server matches GAM's economics AND Google can't quietly starve your non-GAM demand. Two ifs, neither proven. Taboola buying a UK publisher for up to £27 million tells you where the smart money actually goes when open-web growth slows: buy the advertiser relationships, don't wait for a remedy.

Where they part ways

Three real disagreements sit under this.

Does the remedy free anyone, or license the status quo? The Market Analyst sees a capital thaw and new companies getting funded. The Skeptic sees a one-sentence carve-out that makes leaving GAM financially irrational. Both are reading the same document.

Will Google self-correct or route around the order? Adam Heimlich's Microsoft 2001 parallel says senior leadership tires of the fight and competes on merit. But the consensus on that settlement is it was toothless, and open-web competition did the actual work the remedy never forced. And Heimlich concedes his own thesis has a crack in it: product managers with bonuses tied to exchange volume will keep cheating from below no matter what the C-suite decides.

Is the 80% real? Gareth Glaser wants that figure to be true and admits he wishes it had been pulled out at trial. He also runs Gemera, which sells publisher analytics and profits from GAM disintermediation. Informed guess, not a documented number.

What this actually hinges on

One fact decides everything: does AdWords bid meaningfully less into Prebid than into AdX once the remedy is live? If the gap is small, the 80% story has a shot and the capital thaw is real. If the gap is large, and Google's control of its own first-party signals suggests it can be, the remedy is theater and every rational publisher stays put.

The council leans skeptical. Not because the structural fix is badly designed, but because it depends on publishers acting against their own revenue and on Google not using the exact carve-out it negotiated into the document. That's a lot to hope for.

Before building anything: measure the AdWords bid delta the moment AdX-into-Prebid goes live. That single comparison, non-GAM bids versus GAM bids on identical inventory, tells you whether to plan a migration or forget the whole thing.

The Prediction

Prediction: When Google's AdX begins bidding into Prebid under the DOJ remedy, no meaningful wave of publishers will abandon Google Ad Manager within the first year of that integration going live, because AdWords demand delivered outside GAM will price below what it pays inside GAM.

Confidence: Medium. The carve-out is written in, but timing depends on the final order and appeals.

Why: The remedies document explicitly lets AdWords prioritize AdX whenever Google claims better ROI, fraud protection, or privacy, and Google keeps its first-party data signals to itself, so AdWords can bid higher into AdX than into Prebid without ever citing common ownership. That makes leaving GAM a revenue cut for the publisher, and publishers in this market chase yield over principle every time, as the same episode's OpenAI thread shows buyers doing. The opposite outcome, a mass exit, requires 80% of publishers to walk away from money on the strength of an unverified figure that even Gareth Glaser wishes had been proven at trial. Publishers don't do that.

Revisit by 2027-09-21: We're right if, one year after AdX-into-Prebid bidding begins, no independent sell-side ad server has taken a reported double-digit share of former GAM publishers and GAM remains the dominant publisher ad server. We're wrong if a named rival sell-side ad server publicly reports winning a large block of publishers off GAM in that window.

The bolder call is about investor behavior. Nobody funds a serious GAM competitor while that bid delta stays unmeasured. If the delta proves large once AdX-into-Prebid goes live, publisher behavior becomes irrelevant anyway: the capital case for a challenger collapses before it forms.