How to Test AI Max With Google Ads’ New Multi-Campaign Experiments

Google Ads is rolling out multi-campaign AI Max experiments in September 2026. Advertisers can test different budgets and ROI targets across several Search campaigns in one A/B test, while newer experiment controls can preserve brand and location guardrails.

The feature creates a valuable opportunity, but adding more campaigns to one experiment also adds more ways to misread the result. The strongest test changes one strategic lever, protects conversion definitions and evaluates business value instead of conversion volume alone.

What is new in AI Max experiments?

Google announced on August 20, 2026 that advertisers would be able to test budget and ROI-target changes across multiple Search campaigns in a single experiment. The company also said AI Max experiments can run with brand and location controls enabled.

This matters because advertisers previously had to evaluate many scaling changes campaign by campaign. A coordinated experiment can measure a portfolio decision, such as whether loosening target ROAS across a group of campaigns produces profitable incremental volume.

This is separate from the September AI Max transition covered in Google Ads AI Max Is Changing in September 2026. The transition changes eligible campaign settings. The experiment feature helps advertisers measure a proposed strategy before applying it broadly.

When should you use a multi-campaign experiment?

Use it when several campaigns share the same strategic question and comparable economics. Good candidates include campaigns for the same product family, lead type, geography or margin band.

Avoid grouping campaigns simply because they are all Search campaigns. A brand campaign and a high-funnel nonbrand campaign may have different baselines, auction behavior and incremental value. Combining them can produce a blended result that is mathematically correct but operationally useless.

Choose one primary hypothesis

A strong hypothesis names the change, mechanism and business outcome:

Reducing the target ROAS from 500% to 425% across high-margin nonbrand campaigns will increase qualified revenue while keeping contribution margin above the approved threshold.

A weak hypothesis says, “AI Max will improve performance.” That does not specify what is changing or what success means.

Build the campaign group carefully

Use consistent conversion definitions

All campaigns in the test should optimize toward the same primary outcomes or an intentionally comparable value framework. If one campaign counts form submissions and another imports closed sales, the system is not receiving equivalent signals.

Separate campaigns with different economics

Group products by margin or leads by expected value. Revenue can hide a poor result when fulfillment costs, refunds or lead quality differ. The profit-signal framework in this AI bidding guide can help.

Preserve essential controls

If brand exclusions, geographic constraints or text guidelines protect the business, keep them consistent across the control and treatment. Google’s update specifically notes support for brand and location controls in AI Max experiments.

What should the test change?

Change one decision family at a time:

• Budgets while holding ROI targets constant
• ROI targets while holding budgets stable
• AI Max feature activation with approved controls
• A coordinated budget-and-target strategy defined in advance

Changing budgets, targets, landing pages, conversion actions and creative simultaneously may improve performance, but it will not reveal why.

How long should the experiment run?

Do not set duration from a generic two-week rule. Account for conversion lag, weekly seasonality and the number of qualified outcomes. A B2B campaign with a 30-day sales cycle needs a different readout window from an ecommerce campaign with same-day purchases.

Record the decision date, but wait for late conversions before declaring the result. Review whether the treatment changed traffic composition, query mix and lead quality, not only the top-line CPA.

Use a three-level scorecard

Primary business outcome

Measure profit, qualified pipeline, contribution margin or another outcome connected to the campaign’s purpose.

Platform outcome

Track conversion value, cost per qualified action, impression share and marginal efficiency.

Diagnostic outcome

Inspect search terms, landing-page mix, brand versus nonbrand traffic and asset behavior. Diagnostics explain the result but should not replace the primary outcome.

Common experiment mistakes

• Mixing campaigns with incompatible conversion goals
• Using platform revenue when margin differs materially
• Changing campaigns outside the experiment during the test
• Ending the test after one volatile day
• Accepting more conversions without checking whether they are qualified
• Applying a blended winner to every campaign without segment review

Frequently asked questions

Do I need to remove brand controls to test AI Max?

No. Google says newer AI Max experiment capabilities support tests with brand and location controls enabled.

Should brand campaigns be included?

Usually not in the same test as nonbrand growth campaigns. Their intent, incrementality and efficiency baselines are different.

Can I test budget and target ROAS together?

The new planning capability is designed for portfolio budget and bidding changes. However, a simpler test is easier to interpret. Combine changes only when the hypothesis is explicitly about the combined strategy.

Sources

• Google: Make AI Max work with new testing and planning tools, August 20, 2026
• Google: Upgrading legacy Search features to AI Max, updated June 11, 2026