Google Ads adds A/B testing to the Performance Max asset editor

Google Ads now lets Performance Max advertisers turn an asset edit into an A/B test without leaving the asset group editor. Google's help article "About Performance Max optimization experiments: A/B testing assets (Beta)" includes a section, "Set up an experiment from the asset group builder", and Search Engine Roundtable reported the toggle, called "Save and test asset group changes", on 21 October 2026. Once a test starts, the tested asset group is locked in view-only mode until the test ends. A box that applies the winning changes automatically is switched on by default.

What Google added to the Performance Max asset editor

The new panel sits at the bottom of the "Edit assets" page for a Performance Max asset group. Search Engine Roundtable's Barry Schwartz wrote on 21 October 2026 that Google Ads is "possibly rolling out" the toggle, after practitioner Joey Bidner posted a screenshot of it on LinkedIn. The panel text in that screenshot reads: "Measure the performance impact of changes to your headlines, descriptions, images and videos. You won't be able to make changes to your asset group while the experiment is running."

Google's own documentation is the primary source here. The Google Ads help article About Performance Max optimization experiments: A/B testing assets (Beta) (answer 16807329) now describes the builder route step by step. A second help page, "About asset best practice recommendation" (answer 16271597), adds one line: advertisers can "now compare the performance of asset groups by creating an experiment A/B test under 'Save and test asset group changes' in 'Edit assets'".

The steps, as Google documents them

According to help article 16807329, the asset group builder route works like this:

1. Open Campaigns, then select the Performance Max campaign you want to edit.
2. Select Asset groups, then the pencil icon under "Assets".
3. On the "Edit assets" page, change headlines, descriptions, images, sitelinks or other assets.
4. Scroll to the bottom to find the "Save and test asset group changes" panel.
5. Switch on the toggle "Save asset group changes as an A/B test".
6. Review the control arm, treatment arm, experiment name, treatment traffic split, duration and success metrics.
7. Leave or clear the box "Apply these changes to the original asset group if the treatment arm wins", then select Create test.

Two interface details follow from the help page. When the toggle is on, the button at the bottom of the page changes from "Save" to "Create test". And the auto-apply box is on by default, which Google says is "recommended to have on for ease of use". Google adds that the auto-apply option only takes effect when the experiment ends and the results show the treatment arm performed better than the control arm.

The page also mentions a "Save with experiment" option alongside the plain "Save" option. Google describes it as keeping the prior ad version on part of the traffic, so the effect of the edit can be measured against the original settings. The wording on this point in the help article is loose, and the article does not explain whether "Save with experiment" and "Create test" are two names for the same action or two separate buttons.

How this differs from the older Experiments page route

Performance Max asset A/B tests were already possible before this change. The same help article 16807329 still documents the original path through the Experiments page, and comparing the two shows what is new.

This table compares the two documented routes, using only what help article 16807329 states:

The practical change is timing. In the older route, an advertiser had to decide in advance to run a test and go to a separate screen to build it. In the builder route, the test is offered at the moment someone edits creative. That makes it easier to measure routine creative refreshes. It also makes it easier to start a test without meaning to, because the panel shows up during ordinary editing.

How the test splits assets and traffic

Help article 16807329 says an asset comparison test sorts the assets in the group into three categories. Control assets (A) are existing assets used as the reference. Treatment assets (B) are the ones being compared, either other existing assets or new uploads. Common assets are everything not selected for either arm; Google says these "continue to serve normally to 100% of your campaign's traffic", alongside the control and treatment assets in their own traffic splits.

Analysis: the common-asset rule matters when reading results. If only two headlines change and the images, logos and videos stay the same, both arms still serve those shared assets. The test then measures the difference made by the changed assets inside an otherwise identical ad mix, not two fully different creative sets. Joey Bidner made a related point in his LinkedIn post, as quoted by Search Engine Roundtable: the toggle suits testing two messaging angles when you have a complete set of alternative headlines, images and descriptions. That is his opinion, not a Google requirement.

The restrictions Google lists

The "Keep in mind" and troubleshooting sections of help article 16807329 set out the limits. They apply to the test, and several of them decide whether the toggle can be used at all.

The asset group is locked while the test runs

Google states that once an experiment starts, "you can't edit, add, or remove any assets in the asset group that's being tested until the experiment ends." The editing pages for that group switch to view-only. The panel text from the Search Engine Roundtable screenshot says the same thing. If the Experiment Guidance System's 4 to 6 week guidance from the older route is any indication, that could mean a month or more with no creative changes in that group. Google does not state the default duration for the builder route.

Policy review, asset limits and one group per test

New treatment assets go through the standard policy review, and disapproved assets cannot serve in the experiment. Control and treatment assets both count toward the asset group's limits. Each experiment can test assets in a single asset group only. In the troubleshooting list, Google adds that an asset group with more than 15 assets "can be maxed", and suggests removing some assets to get back within the limit.

Campaigns that cannot run the test

Help article 16807329 lists reasons an experiment may fail to create. The campaign has been removed. It uses a shared budget, belongs to a portfolio bidding strategy or uses Smart Bidding Exploration. It already has an experiment running or scheduled over the selected dates. It has features experiments do not support. For campaigns with shared budgets, Google's suggested fix is to remove the shared budget.

What happens when the test ends

Google describes two outcomes. Selecting Apply opens a dialog that updates the original asset group with the changes from the experiment. Selecting End stops the test "without making any changes": new treatment assets are discarded and the asset group reverts to its original state.

Analysis: the default auto-apply box changes how this plays out. With the box ticked, an edit can go live across the whole group without anyone signing it off, as long as Google's results show the treatment won. With the box cleared, a winning treatment still needs someone to select Apply. And if someone selects End, every new asset uploaded for the test is thrown away. Teams should keep copies of new creative outside Google Ads before ending a test.

What Google did not say

The documentation leaves several questions open, and it is better to treat them as unknown than to guess:

• Rollout scope and timing. Search Engine Roundtable called it "possibly rolling out" on 21 October 2026, and Google has not given a date or a list of eligible accounts.
• Country or language limits. Help article 16807329 lists none, but it does not confirm global availability either.
• Default traffic split and duration in the builder panel. The panel shows both fields; the article does not give their default values.
• Which success metrics are available and how a "win" is decided for auto-apply.
• Whether the 15-asset note means 15 assets in total or per asset type. The article gives the number without a breakdown.

The article title still carries the "(Beta)" label, so the details may change.

What to check in your own account before switching it on

Analysis, based on the restrictions above:

• Check the campaign's budget and bidding. A shared budget, a portfolio bid strategy or Smart Bidding Exploration blocks the test.
• Check the calendar. Avoid starting a test on a group you will need to edit soon, such as before a sale where promotion headlines must change mid-flight. The group stays view-only until the test ends.
• Look for other experiments on the same campaign over the same dates, since overlap blocks creation.
• Count assets against the limits, remembering that control and treatment both count.
• Decide on purpose whether to keep the auto-apply box ticked. It is on by default.
• Make the treatment different enough to measure. Changing one headline among many common assets is likely to give a small signal.
• Read the panel before selecting the final button. If it says "Create test", you are starting an experiment, not saving an edit.

What this means for Thai marketers

Help article 16807329 lists no country restriction, so the builder route should apply to Thai Performance Max accounts once the toggle reaches them, though Google has not confirmed availability by market. Analysis: if you plan a Performance Max creative refresh before the 11.11 or 12.12 sales or the year-end season, the asset lock is the main risk. A test started in early November could keep an asset group view-only through the 11.11 and 12.12 sale dates, so any promotion copy you planned to change during that window would stay frozen. Teams that run e-commerce marketing campaigns may prefer to test on a group whose creative can stay fixed for several weeks, and make time-sensitive edits on groups without a running test. Thai advertisers using a shared budget across Performance Max campaigns cannot use the test without removing it first.

FAQ

What is "Save and test asset group changes" in Google Ads?

It is a panel at the bottom of the Performance Max "Edit assets" page that turns your asset edits into an A/B test against the original assets. Google documents it in help article 16807329, and Search Engine Roundtable reported it on 21 October 2026.

Can I edit the asset group while the test is running?

No, the asset group is locked in view-only mode until the experiment ends. Google says you cannot edit, add or remove any assets in the tested group during that time.

Will Google apply the winning assets automatically?

Yes, if you leave the default setting on and the treatment arm wins. The "Apply these changes to the original asset group if the treatment arm wins" box is on by default and only takes effect when the experiment ends with the treatment ahead.

Why can't I create the test on my campaign?

The documented reasons include a shared budget, a portfolio bid strategy, Smart Bidding Exploration, or an overlapping experiment. Google also lists removed campaigns, unsupported features and an asset group with more than 15 assets.

Is this available in Thailand?

Not stated by Google. The help article lists no country restriction, but Search Engine Roundtable describes the toggle as possibly still rolling out, so not every account may see it yet.

The builder toggle makes Performance Max creative testing quicker to start, with the trade-off of a locked asset group and an auto-apply default worth reviewing every time. If you want help planning creative tests around your sale calendar, Relevant Audience's Google Ads team in Bangkok can review your Performance Max setup.