How to Run an AI Max Experiment (and Read It Fairly)

To run an AI Max experiment, go to Experiments in the Campaigns menu. Create a "Campaign features and settings" experiment and choose "AI Max for Search campaigns." Google then splits traffic and budget inside your existing campaign between AI Max off and AI Max on.

The setup takes minutes. Keeping the test fair takes more thought, and some campaigns cannot use this flow at all. This guide is part of our AI ads hub and was checked against Google's documentation on 9 November 2026.

The two ways to test

Google's AI Max experiments page describes two options.

Google says AI Max experiments have a "reduced ramp up time." It says custom trial campaigns "may need a ramp up time, therefore experiments might need to run longer."

Step 1: check whether your campaign qualifies

Google lists six blockers. You cannot create an AI Max experiment through this flow if:

• The campaign has text customization (formerly known as automatically created assets) enabled
• The campaign targets Display network
• The campaign is a part of a Portfolio Bidding Strategy
• The campaign uses Shared Budgets
• The campaign uses Bidding exploration
• The campaign has any active experiments

Google says all but the Display network rule also apply to custom AI Max experiments.

Watch the first blocker. Google's June 2026 update says campaigns using automatically created assets are being upgraded to AI Max "with search term matching and text customization enabled by default." Those campaigns already have text customization on. Check this before you plan a test.

Google also notes: "Turn off the "Optimize your campaign with AI" setting if you can't select that specific campaign for the AI Max test."

Step 2: set it up

These are Google's steps for the Experiments page.

1. Go to Experiments within the Campaigns menu.
2. Select the plus button, then Campaign features and settings.
3. Select AI Max for Search campaigns and click Continue.
4. Click Select campaigns, check the campaigns to include, then select Done.
5. On the "Create an AI Max experiment" page, set the experiment arms, dates, name and auto-apply settings.
6. Select Create.

If you selected only one campaign, you get extra settings. These include "Text customization" and "Final URL expansion" in the "Asset optimization" settings, and "Brand inclusions" and "Brand exclusions."

There is also a faster route. Google's AI Max FAQ says eligible accounts may see a 1-click recommendation that "sets up a 50/50 split test comparing your base Search campaign against an AI Max treatment arm."

Step 3: decide what the trial arm tests

By default, Google says the setup "automatically activates AI Max features including Text Customization, Final URL Expansion, and Search Term Matching in your treatment campaign."

You can deselect Text customization and Final URL expansion for the trial arm. That lets you test search term matching on its own, which is a cleaner question.

Two setup limits matter:

• Ad group changes wait. Google says ad group changes, "such as disabling Search Term Matching or adding URL inclusions," are not supported during initial setup.
• Brand settings hit both arms. Google says brand inclusions or exclusions added during setup "will automatically apply to both your control and treatment arms."

Step 4: keep the test fair

Google's AI Max page does not set a test length. These points come from Google's related pages, plus our own advice where marked.

• Allow for ramp-up. Google's custom experiments page says to "Allow 7-14 days for the treatment arm to stabilize." That page covers custom experiments, but the same caution is sensible here.
• Do not expect equal spend. The same page says a 50/50 split "controls auction eligibility but doesn't guarantee equal impressions or spend."
• Check your conversion volume. The custom experiments page says that for reliable results, base campaigns should meet minimum requirements, "such as more than 100 daily conversions." Smaller accounts can still test, but read the result with more caution.
• Do not edit mid-test. Changing bids, budgets or ads during the test muddies the comparison. This is our advice.
• Pick your success metric before you start. Choose conversions, conversion value or cost per conversion, and decide what result would change your mind. This is our advice.
• Note outside events. Write down sales, holidays and Google rollouts that fall inside the test window.

Step 5: read the results

Google says results are under the "Experiments" tab, with an "Experiment summary" page when you select the experiment campaign.

Look at the trial and control side by side. Then open the search terms and landing pages the trial produced, because a cheaper conversion from the wrong page is not a win. Our AI Max search terms report guide covers what to look for.

The experiment measures AI Max across every placement. It cannot tell you how much of the difference came from AI Overviews or AI Mode, because Google reports those in aggregate.

Step 6: apply it or let it end

Google gives three ways to apply results:

• Select "Apply your experiment changes if results are favorable" during setup.
• Select Apply from the Experiments table after the test ends.
• Turn on AI Max from the Campaign settings page after the test ends.

If you let it end without applying, Google says "all AI Max settings will automatically revert to the original, pre-experiment state." Brand controls added during setup are removed, unless you used legacy brand controls before, in which case those return.

Be careful with auto-apply. Decide whether you trust "favorable" as Google defines it, or whether you want to review search terms and pages first.

Before you run it

Work through our AI Max pre-launch checklist first. Tracking templates and URL exclusions affect the trial arm from its first day.

If your campaign is a Dynamic Search Ads campaign, the path is different. See our AI Max migration plan.

Next step

Choose one campaign that qualifies, write down your success metric and end date, then create the experiment. If you want help designing or reading the test, see our paid search management.

Frequently asked questions