- PMax now takes campaign-level negative keywords, added directly from its own search terms report. [4][2]
- Search themes (up to 50 per asset group) act like broad-match keywords: a real intent lever, not a black box. [3]
- Brand exclusions are a separate, strategic control. Keep PMax off your own brand or a competitor's, across Search, Shopping, and YouTube. [5]
- Channel, placement, and audience reporting mean you can govern PMax like a portfolio, not a wish. [1]
- The controls shipped. The accounts that ignore them are choosing the black box.
For two years, "Performance Max is a black box" was true enough to be a strategy. You handed Google a budget, a pile of assets, and an audience signal, and you hoped. Every agency deck had the same slide about lost control. That slide is now out of date, and the advertisers still reading from it are leaving the single biggest efficiency lever of 2026 untouched.
Google didn't make PMax smarter. It made it governable.
The 2024-2026 updates didn't change the algorithm you're bidding into. They handed back the levers that Search and Shopping always had: query control, placement control, audience control, and the reporting to see what any of it did. PMax is now a broad-match machine you can actually steer.
The controls that actually exist now
Three of them change how you run the campaign day to day.
Search themes are your intent spine. Up to 50 per asset group, they behave like broad-match keywords (directional, not exact) and they replaced the old opaque "search category insights" with something you set on purpose. [3] Cluster them by product range or funnel stage and keep each asset group coherent.
Negative keywords finally work the way you expect. Campaign-level negatives can be added directly in the PMax interface, and (the part that matters) added straight from the search terms report, closing the loop from insight to control without a rep. [2][4]
Brand exclusions are the one people conflate with negatives, and shouldn't.
Brand exclusions decide whether PMax may serve on brand queries at all (yours or a competitor's) across Search, Shopping, and YouTube search inventory, not Display. It's a policy switch.
Negative keywords decide which specific queries the campaign stops matching. It's cost hygiene. Google states plainly they are different mechanisms. Use exclusions for strategy, negatives for waste.
Keep those two levers straight and the wider picture falls into place: this is a campaign type you steer, not one you surrender to.
Performance Max is not a black box anymore: you can now cut waste and improve lead quality by actively managing negative keywords, placements, schedules, devices, demographics, creative assets, audience signals, and URL rules. Mean CEO Blog, Performance Max controls guide 2026 [6]
Why most accounts still behave like it's 2023
The controls shipped quietly, scattered across a dozen release notes, and not one of them switched itself on. The gap between what PMax was and what it now allows is wide, and most accounts are still operating on the left-hand side of it.
No negatives. No search terms. No channel split. Brand traffic swallowed into a blended CPA. "Trust the system."
Campaign negatives, grouped search terms, brand exclusions, per-channel and placement reporting, customer-list exclusions. Governable.
Three habits keep accounts stuck on the old side.
First, one giant PMax campaign doing everything. Brand, non-brand, prospecting, and retargeting blended into a single ROAS that looks healthy precisely because it is quietly harvesting your own brand demand. Split brand out with exclusions and the number gets honest. Usually less flattering, always more useful.
Second, treating the search terms report as a curiosity instead of a workflow. It exists to be mined weekly: grouped queries, thematic negatives, search-theme refinement. The exact loop you would run on a broad-match Search campaign. Accounts that open it once a quarter are funding the queries they would have killed on day one.
Third, reading blended CPA as if it were the truth. Channel-level reporting exists now, and a campaign that looks efficient overall can be dumping budget into cheap Display impressions while Shopping starves for it. Without the channel split, you are optimizing a headline number that hides its own worst input.
Run the controls as a stack, in this order
Layering matters more than any single setting. Apply them hardest-first:
- Account-level negative lists: the non-negotiable blocks (competitor terms, junk generics) that apply everywhere, PMax included.
- Brand exclusions: the strategic call on own-brand and competitor-brand presence, set at campaign level.
- Placement & category exclusions: informed by the network-segmented placement report, for brand safety.
- Search themes + campaign negatives: the ongoing weekly query-hygiene loop, mined from the search terms report.
- First-party audience exclusions: existing customers and low-value visitors removed, so the system hunts new demand.
Read the same layers the other way, by the job each one does and the moment it earns a place in the stack.
| Control | What it governs | Reach for it when |
|---|---|---|
| Search themes | Which queries PMax leans toward | You want directional intent, not a black box [3] |
| Campaign negatives | Which queries it stops matching | A theme leaks waste. Add from the search terms report [2] |
| Brand exclusions | Serving on brand terms at all | Brand belongs in dedicated Search, not PMax [5] |
| Customer-list exclusions | Who it's allowed to reach | You need new-customer, incremental performance [1] |
The real prize is incrementality, not control for its own sake
Control is only interesting because of what it unlocks. Customer-list and website-visitor exclusions mean PMax can be constrained to new customers only. [1] Do that, add brand exclusions so it can't harvest your own demand, and narrow the search themes. What's left is the system working to find people who weren't already coming to you. That is the difference between a campaign that flatters your ROAS with retargeting and one you can actually credit with growth.
The black-box era is over. What replaced it isn't automation you surrender to. It's a portfolio you govern, and the same rules and scripts that keep the rest of the account honest apply here too. Most accounts haven't noticed. That gap is the opportunity.
Sources
- Google · New Performance Max steering and reporting updates coming in 2026
- Google Ads Help · About Performance Max campaigns (controls & reporting)
- Digital Applied · Google Ads Performance Max 2026: campaign guide
- Benly · Performance Max 2026 updates: new controls & features
- Google Ads · New features & announcements: brand exclusions in Performance Max
- Mean CEO Blog · The best Performance Max controls guide for 2026
Frequently asked questions
Can you actually add negative keywords to Performance Max now?
Yes. As of the 2024-2026 updates, campaign-level negative keywords can be added directly in the Performance Max interface, in parity with Search and Shopping, and you can add them straight from the search terms report. Account-level negative lists still apply across PMax too, and remain the most reliable way to block wasteful queries at scale.
What's the difference between brand exclusions and negative keywords in PMax?
They operate on different layers. Brand exclusions are a strategic switch: they control whether PMax can serve on your own brand or on competitor brands at all, across Search, Shopping, and YouTube search inventory (not Display or Discover). Negative keywords are query hygiene: they stop the campaign matching specific wasteful searches. Use brand exclusions for policy, negatives for cost control. They are not the same mechanism.
What are search themes and how many can I use?
Search themes are directional signals (up to 50 per asset group) that behave like broad-match keywords, steering PMax toward related queries without triggering exact matches. Treat them as your intent spine: cluster them by product range, use case, or funnel stage, and align them with the creative in that asset group.
Can Performance Max be used for incrementality rather than just retargeting?
Increasingly, yes. First-party audience exclusions let you exclude specific customer lists and website-visitor segments at campaign level, so PMax can be constrained to new customers only. Combined with brand exclusions and narrow search themes, that strips out the easy branded-retargeting wins and forces the system to find genuinely new demand, which is what makes an incrementality read possible.
Is the search terms report as good as standard Search?
No, and Google doesn't claim it is. The PMax search terms report groups queries by theme with per-group metrics, but it is still not as comprehensive as the full query logs in standard Search. It's enough for weekly query mining and negative-keyword hygiene. Not for forensic analysis.
Why does a single all-in-one PMax campaign make ROAS look better than it is?
Because one campaign covering brand, non-brand, prospecting, and retargeting quietly harvests your own brand demand and blends it into a single flattering ROAS. Split brand out with brand exclusions and the number usually gets less flattering but far more useful. Reading a blended CPA as the truth also hides channel problems, like budget leaking into cheap Display impressions while Shopping starves for it.
How often should I mine the PMax search terms report?
Weekly, as an ongoing query-hygiene loop rather than a quarterly curiosity. Use it to group queries, add thematic negative keywords straight from the report, and refine your search themes, the same loop you would run on a broad-match Search campaign. Accounts that open it once a quarter are funding the exact queries they would have killed on day one.
What order should I apply the PMax controls in?
Hardest-first, as a layered stack. Start with account-level negative lists, then brand exclusions, then placement and category exclusions, then the weekly search-themes-plus-negatives loop, and finally first-party audience exclusions so the system hunts new demand. Layering matters more than any single setting, because the account that skips a layer is optimizing straight through the gap it left.
Mikołaj Salecki
Writes about media, tech, and AI business for people who actually run digital. Former agency lead. Skeptic of frameworks that read better than they perform.
Get the next issue by email.
One letter, once a week. Sharp coverage of media, tech, and AI business. No filler.
