Key takeaways
- Performance Max A/B testing now lets advertisers compare two sets of creative inside one asset group, instead of swapping assets and hoping the numbers move.
- Your current assets act as the control, the new ones act as the treatment, and any common assets keep serving to all traffic.
- You choose the traffic split, so a cautious 80/20 test is possible if you are worried about weaker creative.
- The asset group is locked while the test runs, so plan every change before you press start.
- Judge the winner on qualified leads or revenue, not clicks, and give the test the four to six weeks Google suggests.
Performance Max A/B testing fixes a problem every PPC team knows well. You change a headline set or swap in a new video, performance shifts, and nobody can say for sure whether the creative caused it. Seasonality, bid changes and Google’s own learning all muddy the picture.
In early October 2026, Search Engine Land reported that Google is rolling out A/B testing for Performance Max asset groups more widely, after paid search specialist Bia Camargo spotted it in accounts. The feature lets you run old and new creative side by side within the same asset group, with the traffic split between them, so the comparison is fair.
How Performance Max A/B testing got here
It builds on earlier PMax experiments. Asset tests first arrived for retail campaigns in late 2024, mainly to measure the value of adding images, text or video to feed-only campaigns. Reports in January 2026 showed Google extending asset testing to other campaign types in beta, and this latest rollout brings it into the asset group itself. If you want a refresher on what each asset type does, our Performance Max assets guide covers the basics.
What Google has added to Performance Max A/B testing
Advertisers can now test changes to headlines, descriptions, images and videos within a PMax asset group without replacing the live creative straight away. Each test has three parts:
- Control (A). The assets already running in the asset group.
- Treatment (B). The new or alternative assets you want to try.
- Common assets. Anything you want to keep serving to everyone, whichever arm a user lands in.
You decide how much traffic each arm receives. Google’s Experiment Guidance then estimates how long the test needs to run to reach a reliable result, and Google’s help documentation suggests at least four to six weeks.
How to set up a PMax A/B test
You set the test up from the Experiments section of Google Ads rather than inside the campaign itself. In short:
- Go to Campaigns > Experiments and create a new experiment.
- Choose Assets, then Assets provided by you, then Performance Max.
- Pick the campaign and the asset group you want to test.
- Your existing assets appear as the control. Add or replace assets in the treatment arm.
- Set the traffic split, name the experiment and schedule it.
When the test ends, you have two choices. Applying the experiment adds the treatment assets to the asset group, and you can choose whether to keep the original control assets alongside them. Ending the experiment without applying it discards the treatment assets and leaves the campaign as it was.
Performance Max A/B testing restrictions to plan around
The biggest catch is that the asset group is locked once the experiment starts. You cannot edit, add or remove assets in that asset group until the test finishes. That is deliberate, because any mid-test change would muddy the result, but it means a typo or a policy rejection can cost you weeks.
There are other limits worth knowing before you plan a test:
- Each experiment covers one asset group, so testing several groups means running tests one after another.
- Campaigns using a shared budget, a portfolio bid strategy or Smart Bidding Exploration are not eligible.
- You cannot run an overlapping experiment on the same campaign at the same time.
- Treatment assets still go through policy review, and rejected assets drop out of the test.
- The feature is still labelled beta, so details and availability can change by account.
PPC Geeks’ View
This is a genuinely useful update, but it is easy to run a test that proves nothing. The most common mistake will be testing tiny differences, such as one reworded headline, and expecting a clear winner. PMax mixes assets in thousands of combinations, so small tweaks rarely produce a measurable gap. Test a real idea instead, for example price-led messaging against quality-led messaging, or polished brand imagery against UGC-style content.
The second risk is judging the result on the wrong metric. A treatment that wins on clicks or cheap conversions can still lose on lead quality or profit. For lead generation accounts in particular, the test only means something if your conversion data reflects what happens after the form fill.
An A/B test is only as good as the question behind it. Decide what you are trying to prove, and how you will measure it, before you upload a single asset.
— May Dayang, Digital Marketing Coordinator, PPC Geeks
Creative testing is one of the areas we look at closely in a Google Ads audit, and it is how our Google Ads agency team will use the new tests for clients: one clear hypothesis per asset group, measured against business outcomes.
What advertisers should do next
1. Write a hypothesis for each Performance Max A/B test
Before building anything, write down what you expect to change and why. “Adding product demo videos will lift qualified leads” is a test. “Let’s try some new images” is not.
2. Build a complete treatment set before you start
Because the asset group locks once the test begins, check every headline, description, image and video for spelling, policy issues and landing page fit first. A meaningful test usually needs a full, distinct set of assets in the treatment arm, not one or two swaps.
3. Choose a traffic split that matches your risk
A 50/50 split gathers data fastest. If the asset group drives a large share of revenue, an uneven split such as 80/20 limits the downside while the new creative proves itself.
4. Measure on the metric that matters
Make sure conversion tracking is clean before the test starts. If you can, feed CRM outcomes back into Google Ads through offline conversion imports, so the winning creative is the one that brings in better leads, not just more of them.
5. Give it time and test in sequence
Follow Google’s guidance on duration and resist ending a test early because the first week looks promising. Then roll the winner out and move on to the next asset group or idea. Small, steady tests compound over a quarter.
What Performance Max A/B testing means for your campaigns
Performance Max A/B testing gives advertisers something PMax has lacked from the start: a controlled way to see whether creative changes actually move results. It does not make the campaign type fully transparent, and the locked asset group and eligibility rules mean tests need planning. But for accounts that rely on PMax, it turns creative decisions from guesswork into evidence.
The advertisers who gain most will be those who test bold ideas, measure against revenue or lead quality, and keep a steady testing calendar. If you are not sure whether your PMax creative is pulling its weight, book a free PPC audit and we will show you where the gaps are.
For Google’s own details, see the Google Ads Help guide to Performance Max asset testing.
Frequently asked questions
What is Performance Max A/B testing?
It is an experiment type that splits traffic within a Performance Max asset group between your existing assets (the control) and a new set of assets (the treatment), so you can see which creative performs better.
Which assets can I test in a PMax asset group?
You can test headlines, descriptions, images and videos. Assets you mark as common keep serving to all traffic, so you can isolate the change you want to measure.
Can I edit the asset group during a test?
No. Once the experiment starts, you cannot edit, add or remove assets in the tested asset group until it ends. Check everything carefully before launch.
How long should a PMax asset test run?
Google’s Experiment Guidance estimates the duration for each test, and Google suggests running it for at least four to six weeks so the result is reliable.
Why can’t I create an experiment for my campaign?
Campaigns using a shared budget, a portfolio bid strategy or Smart Bidding Exploration are not eligible, and you cannot overlap experiments on the same campaign. The feature is also still in beta, so it may not be available in every account yet.
