AI Max experiments: budget and ROI tests, brand controls kept on
On 20 August 2026 Google announced three changes to how Search advertisers test AI Max: a single A/B test that compares different budgets and ROI targets across several campaigns, rolling out in September; AI Max experiments that keep brand and location controls switched on so the test reflects the guardrails you actually run; and Performance Planner forecasts for bidding and budget changes that can be applied to campaigns in one click. The brand-control change and the Planner change are documented in Google Ads Help; the multi-campaign budget test is not yet.
| Platform | Google Ads |
|---|---|
| Feature | AI Max experiments |
| Announced | Aug 20, 2026 |
| Status | Rolling out |
| Where available | AI Max experiments with brand and location controls are documented in Google Ads Help. The multi-campaign budget and ROI-target test was announced as rolling out in September 2026 and is not yet described in the help centre. |
| Who is affected | Search campaign owners testing AI Max, especially accounts that rely on brand inclusions or exclusions and accounts using shared budgets or portfolio bidding. |
| Action needed | Optional. If you have been waiting to test AI Max because of brand controls, you can now run the experiment with them on. Check the ineligibility list before planning a test. |
| Sources | Google Ads & Commerce blog: New ways to test and plan with AI Max · Google Ads Help: About AI Max experiments · Google Ads Help: About Performance Planner |
What changed
AI Max for Search is Google’s bundle of three automations for Search campaigns: search term matching, which lets the campaign serve on queries beyond its keywords; text customization, formerly automatically created assets, which generates extra headlines and descriptions; and final URL expansion, which sends traffic to the landing page Google predicts is best. An AI Max experiment splits a campaign into a control arm and a treatment arm with those features on, so you can measure the difference before committing.
The help page gives two ways to start one. From Campaigns › Recommendations, apply the “Turn on AI Max for Search campaigns” recommendation to selected campaigns. Or go to Campaigns › Experiments, click the plus button, choose Campaign features and settings, then AI Max for Search campaigns, pick campaigns, and set the arms, dates, name and auto-apply option on the “Create an AI Max experiment” page.
Three setup details matter. First, the treatment arm turns on text customization, final URL expansion and search term matching automatically; you can deselect the first two under Asset optimization and switch off search term matching per ad group, and those changes only affect the treatment. Second, brand inclusions and exclusions apply to both arms for the whole experiment; afterwards, controls you had before are restored and controls added only for the test are removed. Third, the “Apply your experiment changes if results are favorable” box makes Google roll the treatment into the base campaign on its own when the test ends.
Six kinds of campaign cannot run the experiment: campaigns with text customization already on, campaigns targeting the Display Network, campaigns in a portfolio bid strategy, campaigns with shared budgets, campaigns using bidding exploration, and campaigns with any other active experiment.
The budget and ROI-target test is described in one sentence in the announcement: advertisers can “test different budgets and ROI targets across multiple Search campaigns in a single A/B test,” rolling out in September. Google gave no day, no traffic-split rules and no minimum duration, and the help page has not been updated to cover it.
Performance Planner, the third item, now shows how a bidding or budget-target change may affect existing campaign performance and offers an “Apply suggested changes” button. Its help page lists Search, standard Shopping, Performance Max, Demand Gen and App campaigns as supported, and notes that Display and Video planning ended on 9 March 2026. A campaign is only eligible if its bid strategy has not changed in the previous 10 days.
Why it matters
Brand controls were the main reason careful advertisers stayed out of AI Max. Search term matching and final URL expansion can pull a non-brand campaign onto brand queries, or send brand traffic to a page you did not choose, and until now an experiment could not keep your exclusions in place. With the controls applied to both arms, the test finally measures the configuration you would actually run. That removes the best excuse for not testing.
The ineligibility list is the part most agency accounts will trip on. Shared budgets and portfolio bid strategies are standard in larger accounts, and both block the experiment. Testing AI Max in those accounts means moving a campaign to its own budget and bid strategy first, which is itself a change you would want to let settle before reading results.
Our view: run the first experiment with auto-apply off. Auto-apply is convenient, but “favorable” is Google’s judgment on Google’s metric, and the first test is for learning what the treatment did to query mix and landing pages, not for a hands-free rollout. Keep Performance Planner for what it is: a forecast that should sit next to measured results, never in place of them.
What to do
- Go to Campaigns › Settings and note which candidate campaigns use shared budgets, portfolio bidding, bidding exploration or already have text customization on. Those need restructuring or are out.
- Set brand inclusions or exclusions on the campaign before creating the experiment, so both arms carry the same guardrails for the whole test.
- Create the experiment from Campaigns › Experiments › + › Campaign features and settings › AI Max for Search campaigns. Under Asset optimization, decide deliberately whether text customization and final URL expansion stay on in the treatment.
- Leave “Apply your experiment changes if results are favorable” unchecked for a first test, and set an end date. Read the Experiment summary, then open the search terms report segmented by arm before deciding.
- For the multi-campaign budget and ROI test, look on the Experiments page for a multi-campaign option. If it is not there, it has not reached your account yet.
Still unknown
Google has not published setup details, split rules or a minimum duration for the multi-campaign budget and ROI test, and gave only “September” for its rollout. The Performance Planner help page does not mention AI Max, so how the two interact is undocumented. We have not run this experiment in an account; the search-term and landing-page checks above are how we would read it.
Frequently asked questions
What is an AI Max experiment in Google Ads?
A built-in A/B test that splits a Search campaign into a control arm and a treatment arm with AI Max turned on. You create it from Campaigns › Experiments or from the “Turn on AI Max for Search campaigns” recommendation, and the treatment arm enables text customization, final URL expansion and search term matching by default.
Do brand exclusions apply during an AI Max experiment?
Yes. Google Ads Help says brand inclusions and exclusions apply to both the control and treatment arms for the whole experiment. When it ends, legacy brand controls you had before are restored, and brand controls added only for the experiment are removed.
Which campaigns cannot run an AI Max experiment?
Campaigns that already have text customization on, campaigns targeting the Display Network, campaigns in a portfolio bid strategy, campaigns using shared budgets, campaigns using bidding exploration, and campaigns with any other active experiment.
Is the multi-campaign budget and ROI test available yet?
Google said on 20 August 2026 that it was rolling out in September. As of 14 October 2026 the AI Max experiments help page does not describe it, so check the Experiments page in your account rather than assuming it is there.
Official sources
- Google Ads & Commerce blog: New ways to test and plan with AI Max ↗ · Aug 20, 2026
- Google Ads Help: About AI Max experiments ↗
- Google Ads Help: About Performance Planner ↗
Related coverage
- Demand Gen single-tap image ads on YouTube Shorts and Gmail
- Merchant Center AI performance insights: live in five countries
- Data Strength Uplift: Google Ads’ recovered-conversions metric
Update history
- — Rewritten under the new editorial standard with the full setup path, treatment defaults, the ineligibility list, Performance Planner scope and a practitioner view.
- — Make the documented setup consequences explicit: treatment defaults, brand controls applying to both arms, and optional auto-apply. Keep the budget/ROI rollout unconfirmed.
- — Initial manual coverage under the advertiser-impact standard; original announcement date preserved.
Suggested checks are editorial guidance, not platform requirements. Every claim links to its official source. Report a correction →