Why Performance Max and Advantage+ Are Losing Advertiser Trust

The Pitch the Platforms Made in 2023

Performance Max and Advantage+ were the headline AI ad products of the last cycle. Feed in assets, define a goal, set a budget. The platform decides which placements run, which audiences see what, which creative wins. Google reported over a million advertisers on PMax by late 2024. Meta said Advantage+ Sales Campaigns delivered an average 22% lift in ROAS and grew 70% year over year in Q4 2024, surpassing a $20 billion annual run rate.

The pitch worked on the easy half of the market. DTC brands with a clean conversion event and elastic budgets saw real lift. Sophisticated buyers, agencies running multi-product accounts, and any advertiser with a tight compliance loop saw a campaign type they could not audit.

Two years on, the consensus has shifted. The trade press has gone from cautious optimism to open skepticism.

What the Reporting Actually Shows

The most common complaint about both products is the same: you hand the platform money and you cannot see where it went. Google added a Channel Performance report in 2024 and a Search Partner placements view in February 2026. Search Engine Land's review noted knowing that spend shifted between channels does not let the advertiser reverse the shift, and the absence of channel-level budget controls remains a significant issue. The dashboards got prettier. The steering wheel did not appear.

Meta went the other direction. In March 2025 it removed detailed targeting exclusions from Ads Manager entirely, citing internal testing that showed a 22.6% lower median cost per conversion when advertisers did not use exclusions. Existing campaigns stopped delivering on January 15, 2026, if the exclusion logic was still in place. The control that let a wealth firm avoid showing a retirement product to a 22-year-old, or a beer brand avoid alcohol creative to recovery-community lists, was deprecated.

Search Engine Land's coverage through March 2026 has been sharp. One piece argued PMax and Demand Gen had been mercilessly shopped to advertisers since 2021, while the controls that make them usable arrived years after the sales push began. Another flagged that AI Max accelerated the volume of irrelevant queries with no connection to the advertiser's business. PMax and the illusion of trust, as one headline put it.

The Incentive Problem Nobody Fixed

The structural issue was visible from day one. PMax and Advantage+ optimize for the platform's revenue first and the advertiser's outcome second. The AI layer hides which is happening. When a campaign moves budget from Shopping into Display because Display has higher available inventory and a softer CPM, that move helps the platform clear inventory. Whether it helped the advertiser is a question the advertiser cannot answer without third-party measurement.

Bundling makes the math worse. When high-intent search queries get pooled with weaker placements inside a single bidding pool, the surplus generated by the strong inventory subsidizes the weak. The advertiser sees a blended ROAS number that looks fine and does not see that the strong half of their spend is propping up the rest.

AI is the perfect coating for this. A black box can be defended as proprietary optimization. A black box that ships a 22% lift number on the platform's own dashboards is hard to argue with in a quarterly review.

The Regulated-Industry Cut

We went the other way when we started LeadLord. The wealth managers we talked to had been burned by PMax specifically. The opacity made compliance impossible, because you could not tell which audience saw which variant or which placement drove a given lead. The compliance officer had no audit trail to sign. So we kept the audience targeting explicit and the spend allocation transparent, and we put the AI on the parts where it actually helps: generating compliant variants, killing losing ones, routing qualified leads to a calendar. The marketer keeps the controls. LeadLord puts the AI on the work that scales, not on the decisions that need a paper trail.

Where the Budget Is Moving in 2026

The retreat is from black-box AI in advertising, not from AI in advertising. The shift this year has three shapes.

First, hybrid stacks. An Optmyzr study across 24,702 PMax campaigns found 82% of advertisers were running it alongside Standard Shopping or Search rather than letting it stand alone. PMax becomes a prospecting tier, manual search holds brand and high-intent terms, and the advertiser keeps a campaign type they can audit.

Second, third-party measurement. Server-side attribution platforms, incrementality testing through providers like Haus and Recast, and media-mix models that run independently of the platform's own dashboards. The advertiser stops asking the platform to grade its own homework.

Third, AI as augmentation. Tools that help the marketer write variants, score creative pre-launch, and read post-launch tests faster, without taking targeting or bidding decisions out of the marketer's hands. This is the inverse of the autopilot pitch.

What to Watch

Two things decide whether the platforms can recover.

The first is whether Google and Meta keep adding controls or hold the line on automation. Google's April 2026 release added audience exclusions, budget pacing visibility, and demographic breakdowns inside PMax. Meta has signaled the opposite direction, with the campaign creation experience continuing to collapse toward a goal, a budget, and an image. If those paths diverge, sophisticated advertisers will rotate toward Google for control and lean on Meta for cheap reach.

The second is incrementality. The platforms have spent a decade reporting last-click conversion lift and calling it ROAS. As more advertisers run independent geo-holdouts, the gap between platform-reported lift and real lift becomes a public number. That number decides whether the 22% ROAS claim survives 2027.

The lesson of the last two years is that AI in advertising is real and useful, but the advertiser has to keep the steering wheel. The platforms learned the wrong lesson in 2023. The market is teaching them the right one now.