AI Max Experiments Now Test Budgets and ROI Targets: Your September Rollout Plan

Google announced a fresh round of testing and planning updates for AI Max in Search on August 20, 2026, and the news has been circulating in industry writeups all week. Starting in September, AI Max experiments can test different budgets and ROI targets across multiple Search campaigns in a single A/B test. Performance Planner also gains a preview mode that shows how a bidding or budget change would move existing campaigns, and lets you push those changes with one click (Google Ads announcement).

If you have been holding AI Max at arm’s length because scaling it felt like a leap of faith, this is the release that closes the gap. The new tools cut the setup work, keep your brand and location guardrails intact, and give you a real forecast before you touch a live budget.

What the new AI Max experiments actually test

Before this release, AI Max experiments already gave you a one-click way to split a single Search campaign into a control arm with AI Max off and a trial arm with AI Max on. That model is fast because both arms share the same campaign, so setup errors and sync issues drop out of the picture. It works well for a yes-or-no read on turning AI Max on (Google Ads Help).

The September update pushes that model further in three ways. First, you can now A/B test different budgets and different ROI targets across multiple Search campaigns in one experiment. That means a single test can answer the question every media buyer actually asks, which is not “should I turn AI Max on” but “how far do I scale it before the returns bend.” Second, the experiments now respect specific brand controls and location controls if you have them turned on. You can test AI Max at the campaign level without stripping the guardrails your account depends on. Third, the results roll up across campaigns, so you can see the aggregate impact of a scaling move rather than reading five separate experiment tabs.

Google has been signaling this direction for months. If you migrated during the DSA sunset, our writeup on the Google Ads AI Max migration audit covers the setup you should already have in place before you run these new tests.

How Performance Planner ties into AI Max experiments

Performance Planner is the other half of this release, and the pairing matters. Planner now shows how a bidding change or a budget target change would affect your existing campaign performance, then lets you apply that change in one click. In plain terms, the planner is now a forecast tool for the exact moves you would test inside AI Max experiments.

The workflow that comes out of the pairing is short. You open Performance Planner and model the bidding or budget change you are considering. Planner shows the projected lift or drop against your current baseline. If the projection is plausible, you push the change to an AI Max experiment and confirm the projection with real spend against a control arm. If the projection is not plausible, you kill the idea before you burned a week of budget on it.

Planner was already useful for target CPA and target ROAS forecasting. Our target CPA and target ROAS rename audit walks through the rename that shipped earlier this year and the reporting fields you should double-check before you trust any Planner projection.

Your September rollout plan for AI Max experiments

Five moves this week give you a clean start when the multi-campaign experiments land. None of them require the new features to be live yet. They are the prep work that decides whether your first test is a signal or a noise.

One, list every Search campaign that could be part of a scaling test. Filter to campaigns that have run for at least 30 days, have at least 30 conversions per month, and share a business goal. Campaigns that mix lead-gen and ecommerce goals do not belong in the same experiment, and neither do campaigns on wildly different budgets. Group them into test cohorts before September so you can move on day one.

Two, lock down your brand controls and location controls. The new experiments respect these controls, but only if they are set on the campaign before the test starts. Auditing them now is a five-minute job. Skipping it and setting them mid-experiment breaks the read.

Three, set a real ROI target that reflects your actual margin. AI Max experiments now test different ROI targets across campaigns, which is only useful if your baseline target is already correct. Many accounts are still running with a target ROAS that was set two years ago against a different product mix.

Four, decide on a single budget lever. The experiment can test different budgets, but you still need to pick one budget change to test first. Doubling the budget on a mature campaign and doubling the budget on a new campaign are two different tests and belong in separate experiments.

Five, book a Performance Planner run before the test starts. Get a forecast on your intended budget or bidding change. Save the projection somewhere you can find it two weeks later. When the experiment closes, compare the actual result to the Planner projection. If Planner was off by more than 20 percent, you have a calibration problem that needs a fix before your next test.

Where AI Max experiments still have edges to watch

The new tools are a real improvement, but they do not remove every risk in scaling AI Max. Three edges still cut, and none of them are announced in a product post.

Traffic dilution across cohort campaigns can dampen the read. When you split budget across five Search campaigns in a single test, each one gets a smaller trial arm than a single-campaign experiment would. If your conversion volume is thin, the confidence interval on the aggregate result is wider than it looks. Plan for a longer test window or fewer campaigns per cohort.

Brand queries can quietly reshape the mix. AI Max matches to search terms your keywords do not, and even with brand controls turned on, the boundary between brand and non-brand can drift over the course of a test. Our writeup on AI Max steering controls when CTR and conversions diverge covers the moves that catch the drift before it eats your read.

Bidding target changes inside an experiment can compound. If you change a ROI target and a budget in the same test, you cannot cleanly attribute the result to either change. The new tools let you do it. The new tools also do not stop you from doing it. Discipline still matters, especially given that Google shifted target-based bidding logic in August 2026. Our bidding target optimization audit covers the moves that keep your baseline stable while the test runs.

How AI Max experiments fit your broader account plan

These updates are worth adopting, but they are one tool in a larger stack, not a substitute for account strategy. AI Max is a Search product. Performance Max is a separate product with its own bidding and inventory model. Both matter to the same account, and both feed the same conversion pipeline.

If your account also runs Performance Max, the new AI Max testing tools do not touch that side of the house. Our AI Max September deadline plan covers the ACA and broad match moves that ride alongside these testing tools, and both are worth reading in the same sitting.

If your team is newer to disciplined testing across campaigns, our older A/B testing for lead generation guide covers the fundamentals that keep a test honest. AI Max experiments make the mechanics easier. They do not change what a valid test looks like.

The teams that get real lift from these tools are going to be the ones that treat the September rollout as a scheduling event, not a feature announcement. Book the Performance Planner run before the test. Group the cohort campaigns before the tool goes live. Lock the guardrails ahead of time. Then run the experiment and read the aggregate result honestly.

If your account has more testing questions than time to answer them, our team can run the AI Max experiment plan for you. Our media buying operators have shipped this on live budgets for lead-gen and ecommerce accounts since AI Max first launched. Book a walk-through on our free consultation page and we will map the September rollout to your specific campaign stack. Let’s Grow!