Google adds cross-campaign A/B budget tests to AI Max and one-click Performance Planner changes

Google said on 20 August 2026 that it is introducing a way to test different budgets and ROI targets across multiple Search campaigns inside a single A/B test, and that the capability rolls out in September. The announcement, published on the Google Ads & Commerce Blog, builds on the existing one-click AI Max experiments and arrives alongside two other changes: AI Max experiments can now be run with brand and location controls enabled, and Performance Planner now shows how changes such as bidding or budget targets may affect existing campaign performance, with the suggested changes applicable to campaigns in one click.

What Google announced on 20 August 2026

Three things, and they are easier to judge separately than as a bundle. First, Google is introducing cross-campaign A/B testing of budgets and ROI targets, described as covering multiple Search campaigns in a single test, with a September rollout. Second, new capabilities in AI Max experiments let advertisers run tests with brand and location controls enabled, so those guardrails no longer have to be switched off to test AI Max. Third, Performance Planner gained forecasting of how bidding or budget target changes may affect existing campaign performance, plus one-click application of the suggested changes directly to campaigns. Google's post also points advertisers to its Rethink 2026 virtual event for further guidance.

The table below places each item against its stated timing, using only the dates in Google's own announcements.

Why a cross-campaign test is a different instrument from a single-campaign experiment

A conventional Google Ads experiment splits traffic inside one campaign and compares two versions of it. That answers a question about a setting. It cannot answer a question about allocation, because the budget line it is testing belongs to the campaign it is testing, and money that moves between campaigns is outside the frame.

Google's 20 August 2026 description is of a test that spans multiple Search campaigns in one A/B test, covering budgets and ROI targets. That is the allocation question, which is the one most agency arguments are actually about: whether a target should be looser on the campaign that scales and tighter on the one that converts cheaply, and whether shifting spend across a set of campaigns produces more return than tuning each one alone. Until now the honest way to answer it involved changing several campaigns at once and living with the fact that seasonality, competitor behaviour and the learning period all landed in the same window as the change. A controlled test removes that ambiguity, which is a genuine gain in evidence quality rather than a convenience feature.

What Google's post does not do is describe the mechanics: how traffic or spend is divided, what the unit of comparison is, or what statistical treatment the results receive. Treat the capability as announced and the methodology as unpublished.

Brand and location controls staying on removes the main objection to testing AI Max

Google said on 20 August 2026 that new capabilities in AI Max experiments let advertisers run tests with brand and location controls enabled. That single sentence answers the most common reason serious advertisers refused to run the test at all.

Brand controls and location controls are not preferences. For a business with a trademark to defend or a service area it can physically serve, they are the difference between a test and an uncontrolled spend. An experiment that required switching them off was not comparing AI Max against the current setup; it was comparing AI Max plus the removal of two guardrails against the current setup, and any result it produced was confounded by that removal. Advertisers who declined were making the correct methodological call.

With the controls enabled inside the test, the comparison narrows to the thing being tested. That matters more this month than it would in a quiet quarter, because the window in which testing is still a choice is closing.

Performance Planner can now apply its own forecast, which is a workflow change as much as a feature

Per the 20 August 2026 announcement, Performance Planner now shows how changes such as bidding or budget targets may affect existing campaign performance, and advertisers can apply the suggested changes directly to campaigns in one click.

The forecast half is useful. The one-click apply half deserves a slower look, because it collapses two steps that used to be separate: deciding a forecast is credible, and executing on it. When those steps are separate, somebody usually asks what assumptions the forecast rests on, whether the historical window includes a period that will not repeat, and whether the conversion data feeding it is complete. When they are one button, that question has to be asked deliberately or it does not get asked.

The practical guard is procedural rather than technical. Anyone with edit access can now move budgets and targets on the strength of a forecast in a single action, so teams that share account access should decide who is allowed to use that button before somebody uses it. It is also worth checking the change history after any week where a plan was applied, so an applied forecast is never mistaken later for a deliberate strategy decision. Reliable measurement makes both checks easier, which is one more reason to keep conversion tracking and analytics in good order rather than good enough.

The timing sits days before advertisers lose the choice by default

Google has confirmed that automatically created assets and campaign-level broad match auto-upgrade to AI Max on 1 September 2026, and Google blocked new campaign-level broad match settings on 3 August 2026. The testing tools announced on 20 August 2026 roll out in September.

Read those dates in order and the shape is clear enough. The tools that let an advertiser evaluate AI Max under controlled conditions arrive in the same month that the upgrade happens by default for the affected settings, and Google did not give a date within September for the cross-campaign testing rollout. An advertiser who wants evidence before the change may not have the instrument before the change, and there is no published basis for assuming the tools land on 1 September rather than at the end of the month.

That is not an accusation of bad faith, and the source does not characterise the sequencing at all. It is a scheduling fact worth planning around: build the measurement baseline now, because the pre-change period is the only part of this timeline you control.

What Google's announcement did not say

The post gives no specific date within September for the cross-campaign testing rollout. It does not say whether the rollout is global, staged by market, or limited by account type or spend level. It does not describe how budgets are split inside a cross-campaign test, how long a test should run, or how results are judged significant. It does not say whether the brand and location controls behave identically inside an experiment and outside one. And it says nothing about pricing, eligibility requirements, or minimum campaign counts. Anything you read on those points is inference, including the inference in this article.

What to do in your account this month

• Record a clean baseline now for the campaigns affected by the 1 September 2026 auto-upgrade: cost per acquisition, conversion volume, search term profile and brand share of spend. After the upgrade, a baseline cannot be reconstructed.
• List the campaigns where brand or location controls were the reason you never tested AI Max, because that objection was addressed on 20 August 2026 and the list is now your test queue.
• Check whether campaign-level broad match is in use anywhere, since new settings of it have been blocked since 3 August 2026 and existing ones are in the auto-upgrade path.
• Decide, in writing, who may use Performance Planner's one-click apply, and agree that a plan gets a second reader before it is applied.
• Watch for the cross-campaign testing option to appear during September rather than assuming a date, and design the first test before it arrives so that the test is ready when the tool is.

What this means for Thai marketers

The brand and location control change is the part that matters most in this market, and for a specific reason. Thai-language brand terms are exactly the kind of asset an advertiser cannot afford to leave uncontrolled, particularly where a brand name is also an ordinary Thai word or transliterates loosely, and Bangkok-only or province-level location targeting is standard for clinics, dealerships, property projects and service businesses that can only serve who they can reach. Those were the two guardrails the old experiment flow forced advertisers to drop, which meant a Thai advertiser with either constraint had no defensible way to trial AI Max at all. As of 20 August 2026 a controlled trial in the Thai market is possible, and that is a real change in what a local team can put in front of a client.

The second point is about budget arithmetic. Cross-campaign budget and target testing gets more valuable as the number of campaigns rises and less valuable on a small account with two campaigns and a modest monthly figure in baht, where the split leaves too little on each side to read. Thai accounts run smaller than the US accounts these features are designed against, so the test that proves something in a large account may return noise in a small one. Plan a longer window and fewer, larger tests rather than more of them, and keep your Google Ads structure simple enough that a cross-campaign test has a clean comparison to make.

On availability, the source does not say anything market-specific. There is no published confirmation that Thai accounts get the September rollout on the same schedule as any other market, and no statement that they do not. Check the interface rather than assume either.

FAQ

When exactly does cross-campaign A/B testing arrive?

Google said it rolls out in September but did not give a date within September. The announcement was made on 20 August 2026, so the practical answer is to watch for the option in the interface rather than plan a test around a fixed launch day.

Do I have to turn off brand and location controls to test AI Max now?

No. Google said on 20 August 2026 that new capabilities in AI Max experiments let advertisers run tests with brand and location controls enabled, which removes the requirement that made these tests unusable for brand-sensitive and service-area advertisers.

Is this live in Thailand?

The source did not say. Google's post describes a September rollout without any market-by-market detail, so Thai advertisers should check their own accounts during September instead of relying on a regional assumption.

Does any of this stop the 1 September AI Max upgrade?

No. Google has confirmed that automatically created assets and campaign-level broad match auto-upgrade to AI Max on 1 September 2026, and the announcement of 20 August 2026 does not change that date. The testing tools sit alongside the upgrade rather than replacing it.

Should I use Performance Planner's one-click apply?

Only after you have interrogated the forecast the same way you would if applying it took ten steps. The feature removes friction from execution, not from judgement, and the assumptions behind a plan are unchanged by how quickly it can be actioned.

The short version

Google announced on 20 August 2026 that budgets and ROI targets can be A/B tested across multiple Search campaigns in one test, rolling out in September, that AI Max experiments can now run with brand and location controls enabled, and that Performance Planner can forecast the effect of bidding and budget changes and apply them in one click. The controls change is the substantive one, because it makes a controlled AI Max trial possible for advertisers who correctly refused to run an uncontrolled one. The timing is tight against the 1 September 2026 auto-upgrade, and Google did not give a date within September. If you want a clean baseline recorded and a first test designed before the window closes, Relevant Audience does that work as part of paid search account management.