September 2026 has been AI Max month at Google, whether you signed up for it or not. Dynamic Search Ads auto-migrated. Campaign-level broad match and automatically created assets were pushed to AI Max starting September 1. If you run mid-size or large Search accounts, you've been watching performance shift and trying to figure out how much of it is the migration and how much is normal variance.
That context matters for understanding why Google's August 20 announcement is more significant than its quiet launch suggests.
What Google actually shipped
On August 20, Google announced two changes rolling out through September: a multi-campaign A/B testing feature for AI Max, and a Performance Planner upgrade that forecasts impact before you apply changes.
The testing feature is the one that changes the calculus.
Until now, running an AI Max experiment required disabling brand exclusions and location controls for the duration of the test. That's a real problem — it meant you were comparing AI Max without guardrails against standard Search campaigns with full controls in place. That's not a fair comparison. More importantly, it's not the comparison you actually want: what you need to know is whether AI Max with your brand exclusions enabled outperforms standard campaigns with those same exclusions. The old setup couldn't answer that question.
The new version runs experiments with brand and location controls active from the start. You can test across multiple Search campaigns in a single experiment, set different budget levels or ROI targets for the treatment group, and see how total account performance compares — not just the individual campaign metrics.
Google's announcement also confirmed that the feature supports testing with AI Max's URL expansion and creative flexibility settings enabled, so you're testing the full version of AI Max, not a hobbled variant.
The Performance Planner upgrade
The second piece is a Performance Planner update that now shows projected impact of bidding or budget changes before you apply them. Once the forecast looks reasonable, you can push those changes to your campaigns in one click.
If you've historically found Performance Planner useful but annoying to execute, the one-click apply removes the friction that made it easy to ignore the forecasts. The undo option, available through the Bulk Actions section of Google Ads, means you can reverse an applied change if the numbers don't track.
Google's own performance forecasts should always be treated as directional, not precise. But having a projected view before committing is better than applying budget changes on instinct, particularly during a period when baseline performance is already shifting due to migrations.
Why this matters right now
The September migrations created a measurement problem. Accounts that were auto-migrated to AI Max are now running under different campaign types than they were in August. Pre-migration performance data is no longer a clean baseline. You can't reliably attribute current performance changes to the migration, to seasonality, or to something else entirely.
The multi-campaign experiment framework gives you a controlled path forward. Rather than trying to reverse-engineer what the migration did, you can run a side-by-side test under conditions you define, with your brand and location settings intact.
The accounts most likely to misread their data right now are ones with forced migrations who saw an immediate performance shift and are considering reverting. The issue is that two weeks of post-migration data, especially through a learning period, is not enough to know what's happening. A proper experiment with a four-week minimum window is.
What to do with this
If you have Search campaigns that were recently migrated or that you've been hesitant to test, this week is the right time to set up an experiment. Configure brand and location controls before launching — they need to be in place from the start of the test, not added after the fact.
Give the test at minimum four weeks. The learning phase alone can distort early results, and pulling conclusions from the first two weeks is how accounts end up making structural changes based on noise.
Before touching any campaign budgets, run the scenario through Performance Planner first. The one-click apply makes it easy to act on the forecast, and the undo option gives you a reasonable safety net if the actual results diverge from the projection.
For accounts that were force-migrated and are showing unexpected changes, the experiment data will tell you whether AI Max is the driver. That's worth knowing before you restructure anything.
The free audit at Gromerce maps out which of your current campaigns have enough conversion volume to generate statistically meaningful experiment results. If the volume isn't there, an experiment won't tell you much — and that's useful to know before you commit the testing window.
September gave you the migration. Google also gave you the tools to actually evaluate it. That's the part worth using.
Sources: Google Ads blog, Search Engine Land, Search Engine Journal, August–September 2026
