Performance Max in 2026: The Controls That Finally Made It Auditable

By Ritik Verma, Founder, 7SEA Marketing™ · Last updated: September 2026

7SEA Marketing™ is a Google Partner and Shopify Partner agency with offices in Los Angeles and New Delhi. We run Google Ads for D2C brands, and Performance Max is where most retail spend now lives, so its controls are where most of our audit findings live too.

Performance Max stopped being a black box in 2026. Campaign-level negative keywords (up to 10,000), search term visibility, channel-level reporting, brand exclusions, and proper experiments gave advertisers real levers over what was previously pour-budget-and-pray automation. This guide covers what changed, how to structure PMax now, and the mistakes the new controls expose.

The context for anyone who checked out during the black-box years: PMax runs your products and creative across Search, Shopping, YouTube, Display, Discover, Gmail, Maps, and since late 2025, Waze, from a single campaign. Over a million advertisers run it, and Google keeps steering retail budgets toward it. Resisting it stopped being a viable strategy; auditing it finally became a possible one.

What Changed: The Control Set

Five additions, rolled out across 2025 and 2026, changed the job:

1. Campaign-level negative keywords, up to 10,000

The single biggest complaint about PMax, no way to stop it burning budget on junk queries, is fixed. This closed the largest gap PMax had against standard Search campaigns.

2. Search term visibility

You can now see the queries triggering your ads, which turns negative-keyword work from guessing into pruning. Weekly search-term reviews are back on the checklist.

3. Channel-level and placement reporting

Visibility into where spend actually lands (Search vs Shopping vs YouTube vs Display) ended the era of "trust us, it is working somewhere."

4. Brand exclusions and audience controls

Brand exclusions, refinable per format, stop PMax from claiming cheap branded clicks your brand campaign or organic listing would have won anyway, historically the quietest way PMax flattered its own numbers. Audience exclusions and budget projections arrived alongside in Google's 2026 updates.

5. Experiments

Native A/B testing: PMax against your existing Shopping or Search campaign in a 50/50 split, final URL expansion on versus off, and uplift tests. Covered properly in the experiments section below, because this is how the "is PMax actually better for us" argument finally gets settled with data instead of opinions.

How to Structure PMax in 2026

PMax is configured with signals and assets, not keywords. The 2026 setup that works:

Feed first

For ecommerce, PMax performance is feed quality in disguise: titles front-loaded with what people search, correct identifiers, precise product categories, prices that match the live page. All of it is covered in our Merchant Center guide, and none of it is optional.

Conversion tracking before spend

PMax optimizes on the conversions it can see, so Enhanced Conversions and clean signal are prerequisites, not upgrades. The tracking guide covers the standard.

Creative volume

Max out the asset slots: multiple headlines, descriptions, images, and at least one real video. Two headlines and no video gives the system nothing to mix and nothing to learn from, which is the most common setup failure in audits.

Audience signals as hints, not targets

Customer lists (Customer Match no longer requires a minimum list size), site visitors, and high-intent segments point the machine at your best patterns; it expands from there.

Brand exclusions on day one

Protect your brand terms so PMax's reported ROAS reflects incremental demand, not cannibalized clicks.

Then leave it alone

Learning takes six days to six weeks depending on conversion volume. Pausing, budget-yanking, and daily fiddling during learning is the number one self-inflicted failure; structured patience is part of the setup.

The Experiments Worth Running

Three tests, in priority order for a store:

1. PMax vs standard Shopping, 50/50. The foundational question. Run it four to six weeks at stable budget and judge on your break-even ROAS, not on clicks. Whichever wins gets the budget; the loser becomes the control for the next test.
2. Final URL expansion on vs off. Expansion lets Google send traffic to pages beyond your final URLs. For tight catalogs it often leaks relevance; for broad catalogs it sometimes finds demand. Test rather than debate.
3. Uplift test. Measures whether adding PMax alongside existing campaigns adds incremental conversions or reshuffles the same ones. The honest answer to "is PMax additive for us."

The discipline that makes experiments meaningful: one variable at a time, stable budgets, and a pre-agreed metric. An experiment judged mid-flight by whoever is nervous that week is a coin flip with a dashboard.

AI Max and the September 2026 Migration

The direction of travel is more automation, again. AI Max for Search, Google's suite of AI-powered search features, delivers on average 7% more conversions or conversion value at comparable CPA or ROAS per Google's own published numbers. More concretely for planning: Google is upgrading Dynamic Search Ads into AI Max, with automatic upgrades starting September 2026 for campaigns using automatically created assets with broad match, and the mass DSA migration beginning February 2027.

If you run DSA campaigns, this is a calendar item, not a headline: review what those campaigns do today, decide what should carry into AI Max deliberately, and do it before the automatic migration decides for you. Automation deepening is the constant; the skill of steering it is the job. (Timelines here are Google's announced schedule as of this writing; verify current dates in your account before acting.)

The Mistakes the New Controls Expose

With visibility restored, audits now find the same five problems everywhere:

1. Empty negative lists. The 10,000-slot capability, used zero times. Search term review plus weekly pruning is the new baseline hygiene.
2. Brand terms unexcluded. PMax reporting a glorious ROAS built substantially on clicks your brand already owned.
3. Starved creative. Minimum assets, no video, and a system with nothing to optimize.
4. Learning-phase interference. Budgets cut or campaigns paused mid-learning, resetting the clock and blaming the tool.
5. Feed neglect. Disapprovals, generic titles, and price mismatches quietly capping everything downstream. Fix the feed, and PMax often "starts working" without a single campaign change.

When PMax Is the Wrong Tool

PMax needs conversion volume to learn and a feed to run on. Very low budgets that cannot generate steady weekly conversions, brand-new stores with no conversion history, and lead-gen setups with unverified tracking all feed the machine noise and get noise back. In those cases, start with standard Shopping or tightly scoped Search, build conversion volume and clean signal, then graduate. PMax amplifies inputs; that is its whole nature, in both directions.

Frequently Asked Questions

Does Performance Max support negative keywords now?

Yes. Campaign-level negative keywords arrived with capacity up to 10,000 per campaign, alongside search term visibility so you can see what triggers your ads. Together they closed PMax's biggest historical gap against standard Search campaigns and made weekly query pruning standard practice again.

Is Performance Max better than standard Shopping?

It depends on the account, which is exactly why Google added native experiments. Run a 50/50 PMax versus Shopping split for four to six weeks at stable budget and judge on your break-even ROAS. Well-fed PMax (strong feed, creative volume, clean tracking) usually wins; starved PMax usually loses.

Why is my Performance Max ROAS so high but sales flat?

Check brand exclusions first. Without them, PMax claims cheap branded clicks your organic listings and brand campaigns would have captured anyway, inflating its reported return without adding demand. Exclude brand terms, then re-read performance; the honest number is the one that survives.

How long does Performance Max take to learn?

Six days to six weeks depending on conversion volume, and interference resets the clock. During learning, hold budgets stable and resist edits. The most common PMax failure in audits is not the algorithm; it is a human pausing the campaign in week two and concluding the tool does not work.

What is AI Max and does it replace Performance Max?

AI Max is Google's AI feature suite for Search campaigns, reporting on average 7% more conversions at comparable efficiency, and it absorbs Dynamic Search Ads on Google's announced schedule (automatic upgrades from September 2026, mass migration from February 2027). It runs alongside PMax rather than replacing it; both are the same deepening-automation direction.

Want your PMax account read with the new controls? We run a free Google Ads audit: search terms, brand exclusion status, feed health, creative volume, and experiment plan, with findings ranked by impact. Request your free account audit.