On August 17, 2026, Google changed how target-based bidding behaves for campaigns limited by budget. The algorithm now aims more consistently at whatever Target CPA or Target ROAS the advertiser has set, regardless of whether the campaign had been outperforming that target. A campaign with a $10 Target CPA that was delivering at $5? It'll drift toward $10 unless you recalibrate first.
The industry's initial reaction focused on whether this is a revenue play by Google. That debate misses the structural point. The real question is about who the algorithm works for when incentives diverge.
That Overperformance Was Never Guaranteed
For years, budget-limited campaigns running target-based bidding often delivered below the stated target. Advertisers treated that spread as earned efficiency. It wasn't. It was the algorithm operating with slack. The target was always a willingness-to-pay signal, not a ceiling the system was trying to beat.
Google's August update made that distinction explicit. A system with permission to spend $10 per conversion has no structural incentive to preserve the $5 outcome. The gap becomes room to buy more volume, absorb higher auction costs, or shift traffic mix while remaining technically compliant with the target.
Google built the marketplace, the bidding system, and the measurement layer. It earns more when advertisers spend more. The industry kept treating the platform algorithm as an impartial employee of the advertiser. It was never built to be one.
Three Obligations, One Algorithm
Google, Meta, and every major ad platform sit at the intersection of at least three obligations: return for advertisers, yield for publishers, and growth for shareholders. One algorithm can't cleanly optimize for all three simultaneously. Every public platform is also under the same macro pressure right now: spend at historic levels to win the AI infrastructure race while proving those investments produce returns. Advertising funds that race.
This is the context behind PMax, AI Max, Advantage+, Smart+. The pitch is the same everywhere: hand over your goals, let the algorithm handle bidding and pacing, free your team to "focus on strategy." But a platform-owned system operates within rules the platform writes, using signals the platform controls, buying inventory the platform monetizes, and reporting through measurement the platform provides.
The Agency Problem Behind "Agentic"
The language gets slippery here. The industry is calling this era "agentic advertising." An agent, in the economic sense, acts on behalf of a principal. The principal-agent problem exists when the agent's incentives don't align with the principal's. Google's August bidding change is a textbook illustration: the algorithm isn't malfunctioning. It's doing exactly what it's been told to do, by Google.
A Basis survey from 2026 found that 87% of agency professionals think the traditional agency model is broken or will be within three to five years. 54% said client relationships are more strained. Gartner-linked coverage warns that many agentic AI projects fail due to management issues, unclear business value, and weak controls rather than technical limitations. The blockers aren't model quality. They're governance, interoperability, and operating-model mismatches.
PubMatic's Bill McLaughlin has framed the key opportunity as eliminating inefficiency in pre-execution steps (discovery, negotiation, setup, order management), not in media-buying execution itself. Google's change exposes that: if your governance and target-setting discipline are weak upstream, the algorithm will happily spend to whatever target you left in place.
What This Means for Your Budget-Limited Campaigns
The operational response isn't complicated, but it requires discipline. Leading advertisers are auditing every campaign flagged "Limited by budget," recalibrating tCPA and tROAS targets to recent actuals, and establishing controlled baselines to isolate the change from seasonality. Some are selectively switching campaigns to Maximize Conversions or Maximize Conversion Value where scale matters more than strict efficiency targets.
Google released a Bid Target Adjustment Tool on July 6, 2026 to help with this. Affected campaign types include Search, Shopping, Performance Max, Demand Gen, and Travel. App campaigns, Video Reach, and Video View campaigns aren't in scope.
For B2B SaaS teams, the stakes are higher because the conversion you're optimizing toward matters enormously. If Smart Bidding is optimizing to form fills rather than qualified pipeline, the August change compounds the problem. CRM-connected offline conversion tracking and clear funnel-stage segmentation aren't optional anymore; they're the prerequisite for letting automation work without degrading lead quality.
Governance Is the Product Now
The more platforms automate, the more advertisers need people who can challenge what the machine recommends. Practitioners who catch changes as they happen, forecast the business impact, and design around the risk before it shows up in results. Automate the keystrokes. Don't automate the skepticism.
No agency earns the title of "independent agent" while relaying platform recommendations uncritically, optimizing to platform-reported conversions, or tying its own economics to higher spend. Earning it means maintaining an independent performance baseline, connecting media targets to business economics, and identifying drift before it becomes a quarter-end surprise.
Google's update proved something the industry already knew but kept deferring: automation needs governance. The platforms will keep building more powerful agents inside their own ecosystems. The question that matters is whether anyone on the advertiser's side of the table has the standing, the data, and the incentive to say no.
