Google AI Max Multi-Campaign Testing: What It Means

Google is adding a multi-campaign Search experiment that lets you test different budgets or ROI targets against a control before you commit the whole account. For an Australian advertiser, the point is simple: buy evidence first, then scale. Google says it rolls out from September 2026, so check your own account before planning around it.

Key takeaways

• Google says multi-campaign A/B testing for Search budgets and ROI targets rolls out from September 2026.
• Test before an account-wide change. It gives evidence, not a promise that more spend earns more profit.
• Judge the result on qualified leads and business value, not raw conversions.
• Brand and location controls remove a barrier. They do not remove risk to lead quality, compliance or margin.

What Google changed

On 20 August 2026, Google announced that advertisers will be able to test different budgets and ROI targets across multiple Search campaigns inside a single A/B test, with rollout starting in September 2026 (Google Ads and Commerce Blog, Make AI Max work for your business with new testing and planning tools, 20 August 2026). Google also says AI Max experiments can now run with specific brand and location controls kept on, so you can test the impact without dropping those guardrails. Separately, Google says Performance Planner can model bidding or budget target changes and apply the suggested changes in one click.

Three quick definitions, because these get blurred. A multi-campaign experiment changes a variable across several Search campaigns at once, then compares them to an unchanged group. A control is the group you leave alone, so you have something honest to measure against. A Performance Planner forecast is a model of what might happen, not a record of what did.

Why this matters to an Australian advertiser

Change several campaigns at once with no control and you cannot tell a good decision from a good week. Demand, seasonality and competitors all move. A control gives you an honest baseline. For a business spending real money on Perth trades, professional services or local installs, that is the difference between scaling on evidence and scaling on hope. It does not prove more spend is profitable. It shows whether this change moved qualified outcomes against a fair comparison.

When this test is appropriate

Not every account is ready. You need tracking you trust, a conversion tied to a real business outcome, enough stable activity to read a difference, and a clear guardrail. There is no universal volume threshold, and anyone who quotes one is guessing. If your tracking is shaky, fix that first. A clean experiment on dirty data gives you a confident wrong answer.

How to plan the experiment

Keep it boring and written down. Write the hypothesis as one sentence: what you are changing, and what you expect to move. Select campaigns that are genuinely comparable, then split them into a treatment group and a control. Keep the brand and location settings your account depends on. Choose the decision metric before you start, ideally qualified leads or revenue rather than raw conversions. Then set the test window and the stopping rule, so the result decides, not your patience.

A worked example

Take a hypothetical Perth local-service business running three Search campaigns: urgent callouts, planned installations, and brand. Hypothesis: lifting budgets on the two non-brand campaigns increases booked jobs without hurting cost per qualified lead. Treatment: the higher budget across those two. Control: the same campaigns left at current settings, or a matched holdout, depending on what the experiment allows once live. Brand stays out, so branded demand does not flatter the numbers. Decision metric: qualified enquiries that turn into quoted jobs, not clicks and not raw form fills. No results are claimed here, because the test has not been run. The point is the structure, not a promised number.

Performance Planner is a forecast, not proof

Modelling has a place. Google says Performance Planner can estimate how a bidding or budget change might affect performance. That can shape a hypothesis, but it is not the answer. A forecast is a starting position. The controlled experiment, read against real outcomes, earns the scaling decision. One predicts, the other proves.

Guardrails are not a guarantee

Keeping brand and location controls on during a test is genuinely useful. It removes a reason people avoided testing at all. It does not remove the rest of the risk. Query quality can drift. Compliance still applies. Lead quality can fall while volume rises, and margin can erode while the dashboard looks fine. Guardrails protect the test, not the business decision. That still needs a human reading qualified outcomes.

What to check before testing

1. Confirm the multi-campaign experiment is actually available in your account. Google says rollout starts September 2026, so availability is not guaranteed yet.
2. Verify conversion tracking is accurate and reflects a real business outcome.
3. Pick campaigns that are genuinely comparable for treatment and control.
4. Write a one-sentence hypothesis and choose the decision metric before launch.
5. Keep the brand and location settings your account depends on.
6. Set the test window and a clear stopping rule in advance.
7. Decide the guardrail you will not cross, on spend, cost per qualified lead, or lead quality.

For measurement you can trust, our analytics and reporting work exists to make exactly these decisions readable. If the qualified outcome depends on what happens after the click, conversion rate optimisation is where that gets won.

Test first, then scale

The decision is not scale or hold. It is test, then scale only if qualified outcomes hold up against a fair control. Spend a little to learn before you spend a lot to grow. Our digital advertising team runs experiments against qualified business outcomes, not vanity metrics. So here is the question worth sitting with before the feature lands: if you scaled tomorrow, would you actually know whether it worked?

Frequently asked questions

Is the multi-campaign Search experiment available now?

Google announced it on 20 August 2026 and says rollout starts in September 2026. Availability in a specific Australian account is not confirmed, so check your own account before planning around it.

Will a bigger budget improve profit?

No one can promise that. Google’s feature lets you test a budget or ROI-target change against a control. Whether more spend earns more profit depends on your own qualified outcomes, which the experiment is designed to reveal.

Do brand and location controls make testing safe?

They remove one barrier by letting AI Max experiments run with those settings on. They do not remove risk to query quality, compliance, lead quality or margin. A human still has to read the business outcome.

Sources

• Google Ads and Commerce Blog, Make AI Max work for your business with new testing and planning tools, 20 August 2026, retrieved 2026-08-31, https://blog.google/products/ads-commerce/ai-max-testing-planning-tools/
• Search Engine Journal, Google Ads Launches New Search and AI Max Experimentation Tools, 24 August 2026, retrieved 2026-08-31, https://www.searchenginejournal.com/google-ads-launches-new-search-and-ai-max-experimentation-tools/586549/