Performance Max (PMax) Campaigns in 2026: What Every Advertiser Needs to Know

Whenever something new starts working, everybody wants to talk about it. Have you heard Performance Max mentioned in conversations about marketing and paid advertising?

Here's your guide to staying in the loop and improving those campaigns.

Performance Max is a single Google Ads campaign type that runs one set of assets, budgets, and goals across all seven of Google's channels at once: Search, Shopping, YouTube, Display, Discover, Gmail, and Maps. Google's AI decides which channel serves which person. You supply the goal, the assets, and the audience signals; you do not choose the placement.

That last sentence is why the campaign type was argued about for three years. You hand over the steering wheel and hope.

The argument is so 2024, though. In recent years, Google gave back most of the instrumentation, and the advertisers still losing money on Performance Max are largely those running a 2022 setup on a 2026 product.

‍

What Changed in Performance Max Between 2023 and 2026

For most of its life, Performance Max reported at the campaign level. You could see what it spent and what it returned. You could not see where, on what query, or against which asset.

Here is the current state of the instrumentation:

The bottom four rows arrived together in March 2026, when Google added first-party audience exclusions, budget reporting, expanded audience reporting, and network-segmented placement reports (1). None of them are glamorous. All of them are things advertisers have been asking for since 2022.

The placement visibility landed a month earlier, when PMax data started appearing in the "Where ads showed" report. For this campaign type, the report had been empty since launch (2).

Two limits worth knowing before you get excited about the channel report. Its data only goes back to 6 June 2025, so you cannot use it to re-litigate 2024. And seeing a channel is not the same as steering it: Google's own documentation says you cannot allocate budget by channel, only influence it through assets, signals, and exclusions, and that exclusions "might reduce overall reach" (3).

Visibility, not control. That distinction runs through everything we'll cover below.

‍

How Performance Max Differs From a Standard Campaign

The line that matters is the targeting row. In a Search campaign, a keyword is an instruction. In Performance Max, an audience signal is a suggestion, and Google can ignore it once it thinks it has found something better. Advertisers who read signals as targeting are the ones surprised by their search terms, as reported ninety days later.

‍

Does Performance Max Perform Better Than What You Run Now?

Google's own figure: advertisers who adopt Performance Max see "an average increase of 27% more conversions or value at a similar CPA/ROAS," even when they already run broad match and Smart Bidding (4).

Read the footnote on that number, because Google prints it. It is their own figure, measured over two months in late 2023, about a product that has changed substantially since. Not fraudulent. Not current either.

Independent data tells you adoption rather than lift. Performance Max now takes 67% of Google Shopping ad spend, up from 62% the quarter before. For retail, this is no longer a campaign type. It is the default (5).

For everyone else, treat the 27% as marketing and your own account as evidence. Which is what the next section is for.

‍

The PMax Fit Test: Seven Questions Before You Spend Anything

Run this before you build. Score each question 0, 1, or 2. Fourteen points available.

1. Conversion volume. Does the account already record 30+ conversions a month? (2 = yes; 1 = 15-30; 0 = fewer than 15.)
2. Conversion quality. Can you tell a good conversion from a bad one in your own system, rather than just counting them? (2 = CRM stages flow back to Google; 1 = you can tell manually; 0 = a form fill is a form fill.)
3. Creative depth. Do you have at least 5 images, 1 video, and 5 headlines that apply to every channel? (2 = yes, including video; 1 = images and text only; 0 = you would be making it up.)
4. Budget runway. Can you fund 6 uninterrupted weeks of learning without needing a verdict in week 2? (2 = yes; 1 = 4 weeks; 0 = you need results this month.)
5. Geographic honesty. Is your service area the same as your targeting? (2 = yes; 1 = broadly; 0 = you serve one metro and target a state.)
6. Brand separation. Do you already run a branded search campaign you want protected? (2 = yes, and you will exclude the brand from PMax; 1 = no branded campaign; 0 = yes, and you were not going to exclude it.)
7. Someone to steer it. Is there a named person reading the search terms report monthly? (2 = yes; 1 = quarterly; 0 = nobody.)

11-14: run it. You have the volume, the signal, and the hands.

7-10: run it small, with a stop date: one campaign, a capped budget, and a written review date.

0-6: do not run it yet. Fix the lowest-scoring question first. A Performance Max campaign built on question 2 scoring zero will enthusiastically optimize toward the wrong thing, and the reporting will look fine while it does so.

Question 2 is the one that decides most accounts, and it is the one nobody scores straight.

‍

How to Build a Performance Max Campaign You Can Steer

Six steps. Do them in order; step 1 is load-bearing for everything after it.

‍

1. Define the conversion that pays you, not the one that is easy to count

Set the campaign's conversion goal to the outcome closest to revenue that you can feed back to Google. A booked appointment beats a form fill. A qualified opportunity beats a booked appointment. A closed deal beats it all.

If your CRM can push offline conversions back into Google Ads, do that before you launch, not after. Everything the algorithm does for the next six weeks is downstream of what you told it to want.

Laura Schiele puts the failure plainly: "Performance Max often optimizes for cheap conversions by default" (6). Cheap is what you asked for if cheap is all you defined.

‍

2. Build asset groups by intent, not by product

One asset group per distinct audience-and-offer pairing. Not one per product SKU, and not one giant group holding everything.

The practical test: if two asset groups use the same headlines and images, they are the same asset group. If a headline that is true for group A would be a lie for group B, they are two.

For a fitness client, that means one group for beginners and one for competitive lifters. For a birth center, one for expectant parents and one for the referring providers. Same brand, different sentence.

‍

3. Fill every asset slot, and include video

Give each group the maximum you can produce without padding: headlines, long headlines, descriptions, logos, images in every ratio, and at least one video. Google will auto-generate a video if you skip it, and the auto-generated one is a slideshow of your stills with a soundtrack.

Make the video real. It is the only asset in the group YouTube can use properly, and a slideshow is what a slideshow looks like on a channel people go to for video.

Asset fatigue is real and predictable. We refresh paid creative monthly as the standing cadence and swap visuals every 6-8 weeks at the absolute minimum. We build the next batch before the current one decays, rather than after leads drop.

‍

4. Feed the signals: search themes, audience signals, first-party data

Three inputs, in descending order of how much Google listens.

• Search themes—up to 50 per asset group, raised from 25 in 2025. Use the queries you already know, converted, taken from your Search campaign's search terms report, not from imagination.
• Audience signals. Customer Match lists, site visitors, high-value converters. Seed with people who took the valuable action (booked, bought, qualified), not everyone who ever filled in a form.
• First-party exclusions. New in 2026. Exclude your existing customer list if the goal is acquisition, so you stop paying to reach people who already bought.

Signals are direction, not fences, which is why step 5 exists.

‍

5. Put the fences up before launch, not after the first bad month

Four exclusions, all available now, all skipped by most accounts:

• Negative keywords, up to 10,000 per campaign. Load your existing Search campaign's negative list on day one.
• Brand exclusions, if you run a branded search campaign. Without them, PMax will happily take credit for the people already typing your name.
• Location targeting is set to "presence," not "presence or interest," unless you serve people who are merely curious about your city.
• Placement exclusions, set at the account level, for the Display and YouTube inventory where accidental clicks convert at a rate that appears enthusiastic and inflate your lead count.

Google's documentation is honest about the fact that exclusions can reduce reach. For a lead-gen account, reduced reach is usually the point.

‍

6. Set the review cadence and the kill criteria before you launch

Write down, before the campaign goes live, what would make you turn it off. Ours, applied across the paid accounts we run:

• A new customer is costing more than twice what you budgeted, seven days running.
• Fewer than 5 people in 1,000 are clicking, after three different creatives have had a turn.
• It is taking more than 18 months for a customer to pay back what you spent to win them, and they don't stay long enough to make the wait worthwhile.

That third one is the one people skip, and it decides whether the other two even matter. An 18-month payback period is fine if your clients stay for 9 years. It is fatal if they stay for seven months.

These are our working defaults, not industry law, and they are tuned per account. The value is not the specific threshold. It is that somebody wrote a number down while calm, so the decision in week 9 is a check rather than an argument.

Then the monthly job: read the search terms report, add negatives, read the channel report, and check whether the spend mix matches the intent you thought you were buying. Thirty minutes. It is the difference between an AI campaign and an unsupervised one.

‍

The lead-gen problem nobody writes about

Almost everything published about Performance Max is written for e-commerce, where a conversion is a purchase, and the value is unambiguous. Our client roster is mostly not that. We have executive coaching businesses, wellness and financial institutions, hospitality ventures, and nonprofits, where a conversion is a form, and its value ranges from a signed contract to a bot.

That difference changes the setup, not the campaign type.

Performance Max optimizes toward whatever you call a conversion. Feed it undifferentiated form fills, and it will find you the cheapest possible form fills, which is exactly what it was asked to do and exactly what you do not want. Display and YouTube inventory make this worse because accidental clicks convert at a rate that appears enthusiastic.

The fix is the same shape every time. It is mostly steps 1 and 5 above: define the conversion at the qualified stage, push CRM data back into Google, seed audience signals with people who booked rather than people who clicked, harden the form itself (validation, disqualifying questions, blocking free email domains), and put the placement exclusions up before launch.

What doesn't fix it: changing the bid strategy, adding budget, or adding assets. All three are reliably useless on lead quality, and that matches what we see. Lead quality is a definition problem wearing a targeting costume.

For lead-gen accounts with low conversion volume, a well-built Search campaign with real keywords still often beats Performance Max, so PMax should be a test, not a migration.

The question is never "should we be on PMax?" It is "what is the constraint in this account," and sometimes the answer might be a landing page or a new asset. It's all about testing based on an understanding of what's valuable for the business.

‍

Where This Could Be Wrong

Three caveats you need to consider for your brand.

Cannibalization is overstated. Search and PMax overlap on keywords in 91.45% of accounts, the number everyone quotes as proof that PMax is eating your Search traffic. The same study found no measurable difference in 74.92% of cases. Manage the overlap with brand exclusions and negatives, but do not wall PMax off from every query Search already owns. Most of that walling changes nothing, and each exclusion costs you reach.

The fit test is our method, not Google's rules. So a 10 and an 11 are not different in kind. Read which question scored lowest, not the total.

And we are biased toward caution. We run paid media mostly for lead-gen and mission-driven clients on modest budgets, where one bad month is felt immediately. If you are retail with real conversion volume, discount our caution accordingly.

What we run either way: test small, define the conversion properly, and never migrate an account that is already working.

‍

What Good Looks Like at 90 Days

Not a ROAS number. What we see will be completely different from what you can see, and any blog that hands you one is guessing about your margins.

Four checks instead:

1. The search terms report contains queries you would have bid on. If it is full of terms you would never buy, your negatives and search themes are underbuilt.
2. The channel report shows a mix you can defend. If 80% of the budget went to Display on a lead-gen account, that is a finding, not a result.
3. Lead quality held. Same qualification rate as your other channels, measured in your CRM rather than in Google Ads.
4. Your Search campaign did not quietly shrink. Check impression share on your branded and core non-branded terms before and after.

Fail any, and the answer is usually step 1, not a bid strategy.

‍

Run The Fit Test Before You Build Anything.

Performance Max in 2026 is a good campaign type with defaults that suit retail and a control surface most advertisers never touch. The gap between those two facts is where the money goes.

Score the seven questions. If you land above 11, build it properly with steps 1 through 6. If you land below 7, you have found something more valuable than a new campaign type: the actual constraint in your account.

If you want a second read on the score or advice from our team, schedule a complimentary call, and we can guide you through your test.