Google Ads recently changed how target-based bidding behaves on budget-limited campaigns. The algorithm now bids more aggressively to track closer to whatever target ROAS or CPA you’ve set, regardless of historical performance. Google flagged the change in advance and was transparent about the expected outcome: a campaign with a $10 target CPA that has been delivering at $5 will move closer to $10, unless the advertiser changes the target first.
The industry’s first reaction was to debate whether this is a cash grab. The more consequential question is why the buy side remains so smitten with the idea of handing maximum control to autonomous buying systems owned by companies whose economic interests are not its own.
The efficiency was never yours
For years, advertisers could set a target that gave the algorithm room to operate and still benefit when the system found conversions more efficiently. That overperformance created value for the advertiser. The change makes clear that the entered target is not a ceiling that the platform will try to beat. Rather, it is a price the advertiser has said it is willing to pay. That distinction matters. A system with permission to spend $10 has less incentive to preserve the $5 outcome. The available efficiency becomes room to buy more volume, shift traffic or absorb higher costs while remaining technically on target.
Nothing about that is irrational. Google built the marketplace, the bidding system and the measurement environment. It earns more when advertisers spend more, and it genuinely believes consistency in platform outcomes is a win for advertisers. The problem is that the industry keeps describing the platform algorithm as if it were an impartial employee of the advertiser. It never was, and it was never built to be.
This is not just a Google story
Google, Meta and every major platform sit at the intersection of at least three obligations: return for advertisers, yield for publishers and growth for shareholders. One algorithm cannot cleanly optimize for all three at once.
Every public platform is also under the same basic pressure: Spend at historic levels to win the AI race while proving those investments can produce equally historic returns.
Meta’s latest quarter shows how intense that pressure has become. Revenue rose 28% year over year, yet costs and expenses rose by 55%. Free cash flow fell by 91% as AI infrastructure spending consumed nearly everything the business generated. Meta also raised its full-year capex range again to $130-145 billion. The stock fell roughly 9% after the report. These are the results when platforms are under a mandate to turn AI capability into commercial yield. Advertising is not adjacent to that mandate. Advertising is what funds it.
Meanwhile, platforms need automation to create more value. But platforms also need to capture more of that value for themselves. So advertisers should stop assuming platform automation is always for their benefit.
Whose agent is it?
This reality creates an awkward moment for the industry’s current fixation: fully agentic, autonomous media buying.
Whether it’s Google’s PMax and AI Max, Meta’s Advantage+ or TikTok’s Smart+, the messaging is the same. Hand the platform your goals and let the algorithm handle bidding, pacing and optimization. Free your team to “focus on strategy.”
But whose agent is it really?
A platform-owned system operates within rules the platform writes, using signals the platform controls, buying inventory the platform monetizes and reporting performance through measurement the platform provides. When advertiser efficiency and platform economics diverge, there’s no reason to expect the system to choose the advertiser.
Google’s recent bidding change demonstrates the problem in plain sight. The algorithm is not malfunctioning; it is doing exactly what it is being asked to do.
This is the principal-agent problem of the AI era: Advertisers and their agencies are being handed increasingly powerful agents while forgetting who the principal is.
Expertise is being devalued at exactly the wrong moment
The advertising industry has spent years treating practitioner expertise as friction to be engineered out. Platform interfaces have grown simpler while the decision systems underneath them have grown more complex and less observable. That combination is convenient right up until the platform changes what the machine is optimizing toward.
A media buying practice is valuable because someone has to decide whether the bid, the target and the machine are still serving the advertiser’s goal. That requires practitioners who know the platforms well enough to catch changes as they happen, forecast the business impact and design around the risk before it shows up in results. Often, it means spotting the changes that were never announced at all.
The more platforms automate, the more advertisers need people empowered to challenge what the machine recommends. Automate the keystrokes. Do not automate away the skepticism.
Agencies have to earn this, not just claim it
No agency can claim to be the advertiser’s independent agent while relaying platform recommendations, optimizing to platform-reported conversions or attaching its own economics to higher spend.
Earning the role means doing what a platform cannot credibly do for itself: maintain an independent performance baseline, connect media targets to business economics, compare outcomes across platforms, disclose incentives and identify drift before it becomes a quarter-end surprise.
Google’s update is proof that automation needs governance. The platforms will keep building more powerful agents inside their own ecosystems, and those agents will keep getting better. The risk is that the advertiser is underrepresented.
An arms race won’t fix that. Machines bidding against machines, burning compute on every round that someone eventually pays for, is a losing proposition for the advertiser funding both sides of the fight. What advertisers need on their side of the ledger is judgment: someone who knows what the target was supposed to mean, notices when the definition of success quietly changes and has the standing to say no.
For advertisers and their agencies, there are only two messages on offer right now: media buying still takes expertise, or the platforms have solved it for you. Only one of them leaves someone in the room to critically evaluate the next change before it hits your bottom line. “Data-Driven Thinking” is written by members of the media community and contains fresh ideas on the digital revolution in media.
Follow Goodway Group and AdExchanger on LinkedIn.