Performance Max Gets Real Controls: Campaign-Level Negatives, Brand Exclusions, and Channel Reporting Are Now Live

Performance Max spent its first three years as a machine you could point at a goal but could not meaningfully steer. Advertisers handed over budget, creative, and conversion targets, and the algorithm decided everything else: which queries to enter, which channels to buy, which assets to show, which audiences to prioritise. The feedback loop was thin. You could see aggregate numbers. You could not see much underneath them.

That has changed materially. Campaign-level negative keywords rolled out to all advertisers in January 2025 after a beta announced at DMEXCO the previous September. Brand exclusions became granular enough to apply differently to Search text ads and Shopping in the same release. Channel-level reporting, which once showed only blended totals, now breaks down performance across all seven networks. Asset ratings moved past the coarse Best/Good/Low labels to full per-asset performance metrics. These are not incremental refinements. For anyone running PMax accounts seriously, they represent a qualitative shift in what is actually manageable.

This piece works through each control in turn, covers what is still not in your hands, and closes with a practical audit checklist for existing campaigns.

The Controls That Were Missing and Why Advertisers Pushed Back

The criticism of Performance Max was never really about the automation itself. Most practitioners running at scale were already comfortable with Smart Bidding, which is also a black box in important respects. The problem was that PMax operated at a level of opacity that made it genuinely difficult to diagnose problems, protect brand equity, or trust that budget was going anywhere sensible.

Three complaints dominated the feedback from 2021 through 2024. First: PMax would spend on branded search queries, cannibalising existing brand campaigns at a higher CPA and muddying attribution. Second: there was no way to exclude keywords at campaign level, so any irrelevant or damaging query pattern had to be escalated to an account-level negative list, which affected all campaigns. Third: the only channel-level visibility was inferential, based on asset group performance or third-party tools, rather than reported directly. You were optimising based on output metrics for a system whose inputs you could not examine.

The product team at Google heard this consistently. The January 2025 announcement and the subsequent 2026 updates address all three complaints, though some with more completeness than others.

Campaign-Level Negative Keywords: What Rolled Out and When

Negative keywords for Performance Max arrived in beta at DMEXCO in September 2024. Per Google's announcement on 23 January 2025, they were rolling out to all advertisers within weeks. By 2026, full negative keyword support was confirmed at both campaign and account level, with broad match, phrase match, and exact match negatives all available.

The practical implementation is cleaner than some practitioners expected. The search terms report now includes a one-click workflow to add a term directly as a negative, matching the experience in standard search campaigns. You can build campaign-level negative lists as you would anywhere else in the account, and those lists apply only to the PMax campaign they are attached to, leaving other campaigns unaffected.

The account-level negative list remains relevant for brand-wide exclusions: competitor terms you never want any campaign entering, regulatory-risk queries, and terms that are irrelevant to your entire product set. Campaign-level negatives handle the more specific exclusions: seasonal terms that do not apply to this campaign, product categories this specific campaign does not cover, and queries with a demonstrated low-conversion pattern that is specific to PMax rather than your search campaigns.

The shift from account-only to campaign-level control also changes the workflow for launching new PMax campaigns alongside existing search campaigns. Previously, the safest approach was to add PMax-specific exclusions at account level and accept the collateral effect on search campaigns. Now you can isolate those exclusions properly.

Brand Exclusions: The Difference Between Search Text Ads and Shopping

The brand exclusion update in January 2025 addressed a specific conflict that was particularly acute for retail advertisers running product feeds. PMax was bidding on branded search queries in two ways simultaneously: triggering Search text ads for someone searching for the brand name, and serving Shopping ads for the same query. For many advertisers, the Shopping exposure was welcome, because branded Shopping queries often close at high ROAS, while the branded Search text ad exposure represented wasteful spend already covered by a dedicated brand campaign.

The old exclusion model did not distinguish between these two surfaces. If you excluded a brand from PMax, you excluded it from both Search text ads and Shopping. The January 2025 update changed this: brand exclusions for retail advertisers with product feeds can now be applied only to Search text ads, leaving Shopping unaffected.

For accounts that were managing this conflict by running separate PMax campaigns split by feed and non-feed, this change removes the need for that workaround in many cases. You can now run a single PMax campaign with a product feed, exclude branded terms from text ads, and retain branded Shopping coverage.

Wordstream noted in January 2026 that brand exclusions had expanded further to cover all Search and Performance Max campaigns, and flagged an important consequence: advertisers who do not actively configure brand exclusions remain exposed to competitor brand terms triggering their PMax ads. The system does not apply exclusions by default. You have to act.

Channel-Level Reporting: Finally Seeing Where Budget Goes

The reporting change is the one that should change day-to-day workflow most immediately for anyone who manages PMax at any scale. PMax now breaks down performance across all seven networks: Search, Shopping, Display, YouTube, Discover, Gmail, and Maps. The report shows impressions, clicks, cost, and conversions per channel.

Per Benly's comparison of PMax 2025 versus 2026 controls, channel reporting moved from "not available" to "complete breakdown by network." That is not a small change. For most accounts, the blended ROAS figure for a PMax campaign conceals a wide range of per-channel efficiency. Search and Shopping tend to perform near or above account targets. Display and Gmail tend to perform significantly below. Discover and YouTube vary considerably by account type. Without channel-level data, diagnosing a declining ROAS figure required inference and testing. With it, you can see which channel is pulling the average down.

Most advertisers running the new 2026 PMax controls are seeing 15 to 25 per cent improved ROAS compared with previous PMax versions, based on Benly's early-adopter data.

Benly, PMax 2026 updates (January and March 2026)

The important caveat is that seeing channel-level data does not mean you can act on it directly. Budget allocation between channels remains under algorithmic control. You cannot instruct PMax to spend 80 per cent on Search and Shopping and reduce Display to 20 per cent. What you can do is use the channel data to set appropriate expectations with clients and stakeholders, to flag campaigns where Display or Gmail is consuming a disproportionate share without converting, and to make an informed case for restructuring or for running separate campaign types where the channel mix is genuinely misaligned with objectives.