Google Ads

Performance Max Lets You Save Asset Group Edits as an A/B Test

Google Ads has updated Performance Max with a feature advertisers have wanted since the campaign type launched: the ability to save asset group edits as an A/B test. Instead of setting up a separate experiment, you can now turn an edit into a controlled test directly from the asset group builder, with the traffic split and duration set in the same place.

How the new test flow works

Advertisers can access the feature through the Edit assets page, where a new panel labelled "Save and test asset group changes" allows toggling the "Save asset group changes as an A/B test" option. The toggle opens a view where you can review the control and treatment arms, set the experiment name, choose the treatment traffic split, and set the duration.

The change simplifies the process by removing the need for separate setup. Before this, testing a new set of assets in Performance Max meant building the experiment apart from the edit itself, which added steps and discouraged small tests. Now the test is part of saving the edit, which makes it far more likely that advertisers will actually test instead of just changing assets and hoping.

The update also includes adjustments for feed-only experiment pages, where control and treatment are framed as feed-only asset groups. Those require an existing campaign with at least one paused feed-only asset group. Google documents the feature on its help page "About Performance Max optimization experiments: A/B testing assets (Beta)," which signals that the capability is still rolling out and may not be in every account yet.

Why testing inside Performance Max has been so hard

Performance Max is Google's most automated campaign type: one campaign can serve across Search, Shopping, YouTube, Gmail, Discover and the Display Network, with the system choosing which assets to show to whom. That automation is the point of the product, but it makes creative testing difficult. When the system controls serving, an advertiser cannot simply show version A to half the audience and version B to the other half by hand.

The usual workaround has been to change the assets and watch what happens — a before-and-after comparison that is vulnerable to everything else that changed in the meantime, from seasonality to competitor behaviour to Google's own system updates. A structured A/B test with a control arm and a treatment arm is the only way to know whether the new assets actually caused the change, and it is the standard that serious testing demands.

That is why this feature matters more than its modest interface suggests. It lowers the cost of doing a proper test from a separate project to a checkbox at the moment of editing. The advertisers who will benefit most are the ones who iterate on creative often: ecommerce brands rotating seasonal imagery, lead-generation advertisers testing new headlines, and agencies managing many asset groups across many accounts.

How to use it well

The mechanics are simple, but the discipline around them is what makes a test useful. Change one thing at a time: if the treatment arm swaps the images, the headlines and the audience signal all at once, a win tells you nothing about which change caused it. Set the traffic split with the campaign's volume in mind — a 50/50 split on a low-volume campaign will take a long time to reach a readable result, and a short duration on a slow converter will end before the data means anything.

It is also worth remembering what Performance Max itself needs to learn. Google's own troubleshooting guidance now says a campaign needs at least 15 conversions over 30 days for stable performance. Splitting traffic between two arms halves the learning signal each arm receives, so a campaign that is already short on conversions will learn even more slowly during a test. Test when the campaign has enough volume to support two arms, and give the test enough time to clear the noise.

Finally, treat the result as a decision, not a decoration. Decide before the test starts what outcome would make you keep the new assets, what would make you revert, and how long you will wait. A test with no pre-set decision rule is just a change with extra steps, and the whole point of the new toggle is to make the disciplined version as easy as the sloppy one.

Sources

Sources: PPC News Feed, November 3, 2026; Google Ads Help, "About Performance Max optimization experiments: A/B testing assets (Beta)."

Related reading on this site: SEO Split Testing in 2027: How to Prove a Change Actually Worked, Google Ads Tests New Asset Activity Section With Weekly AI Insights, Google Favours High Ad Strength Ads, But Lower-Rated Ads Convert Better, Google AI Mode Multiturn Ads Are Now Live in Follow-Up Queries, How to Measure the Full Value of Paid Media Beyond ROAS, and Google Is Testing AI-Generated Summaries Above and Within Search Ads.

More from Linkbaits

Turn this playbook into links

Linkbaits helps you create content that earns citations and backlinks. Start with the toolkit and build your first linkable asset this week.

Start free trial