Key takeaway: Google Ads has documented a one-click setup for AI Max experiments in its frequently asked questions page for AI Max search campaigns.
Key Takeaway
Google Ads has documented a one-click setup for AI Max experiments in its frequently asked questions page for AI Max search campaigns. If an account is eligible, a recommendation may appear on the Recommendations page and automatically create a 50/50 test comparing the existing Search campaign with an AI Max treatment arm. Separately, reporting on Google Ads operations warns that meeting a tCPA or tROAS target does not necessarily mean that revenue, profit, or overall business growth is increasing.
For advertising platform teams, the practical implication is twofold: test the AI Max configuration with a clearly defined control, and verify that the bidding target still reflects the business outcome that matters. The available reporting confirms the documented experiment workflow and the measurement risk. It does not establish that AI Max is available to every account or that it produces a consistent performance improvement.
What the one-click AI Max experiment changes
For eligible accounts, selecting “Create 1-click experiment automatically” sets up a 50/50 split test between the base Search campaign and an AI Max treatment arm. Advertisers can also create an AI Max experiment through the Experiments tab and apply more granular controls at the ad-group level. The documented controls allow Search term matching or Asset optimization to be enabled or disabled for individual ad groups in the treatment arm, while the control arm retains the base campaign settings.
This gives operators a defined checklist before launch: confirm whether the recommendation is present, identify the control and treatment campaigns, and record which ad groups and settings are included. The source material does not provide eligibility criteria, account coverage, test duration, or a universal launch date. The documentation update therefore should not be treated as proof that the feature is available to all Google Ads advertisers.
Why Google Ads targets can hide weak growth
Smart Bidding optimizes toward the target it receives. An account can therefore meet its tCPA or tROAS while the underlying business remains flat if the target no longer matches current economics or if the conversion signal is a weak proxy for revenue. The reported issues include targets that were not revisited after changes in costs, margins, shipping, product mix, or customer behavior; conversions concentrated among returning customers or branded searches; and lead actions such as form fills, downloads, or booked calls that never become sales.
Strict efficiency targets can also restrict profitable reach. A low CPA or high ROAS requirement may cause the system to avoid more expensive auctions that could have generated incremental volume. Operators should compare targets with current margins, separate new from returning customers, examine the contribution of branded campaigns, and check whether conversion data is connected to closed revenue. They should also inspect what happens after the click, including lead follow-up, stock availability, and fulfillment. Reaching an advertising target demonstrates compliance with a platform instruction, not business growth by itself.
Method and Limitations
Teams can evaluate the AI Max test by documenting the control and treatment settings, the ad groups included, and the actual business outcomes alongside platform metrics. Revenue, profit, and new-customer contribution should be reviewed with CPA, ROAS, and conversion volume. Lead-generation advertisers should also verify how submitted forms or booked calls progress to closed business.
The public material does not confirm universal access to one-click experiments or provide standardized AI Max performance results. The Smart Bidding concerns are operational checks rather than a claim that every account has the same problem. Results will depend on the account’s experiment configuration, conversion definitions, and downstream business data.