My AI Max experiment didn't change my mind
I recently tested Google's AI Max for Search. The experience didn't change how I think about it. The capability is worth understanding, but my initial testing didn't build a strong case for broader adoption.
I'm not arguing that AI Max has no future. My view is that a new capability should have to show it improves on an established, well-managed program before it gets more budget. In my test, it didn't.
What I was trying to learn
The question was whether AI Max can create meaningful incremental business value beyond what a mature paid search program already captures. That it works mechanically was never in doubt. Whether it helps the business was the open question.
What gave me pause
Expanded matching isn't necessarily incremental reach. If automation starts serving searches that an existing campaign already captures efficiently, the question becomes what additional value it creates. I wanted to know whether we were reaching new demand or redistributing demand we already had. More traffic doesn't answer that, and neither does eligibility for more auctions.
The results weren't compelling enough to justify expanding adoption. For me, that's the point of testing: determining whether a new capability improves the business, not simply whether it works.
Testing also reminded me to look past topline performance. How automation interacts with your existing campaign structure is part of deciding whether it's ready to scale.
Why I didn't enable every feature
AI Max bundles search term matching, text customization and final URL expansion. I evaluated them selectively, because each one touches something I'm responsible for.
- Text customization affects brand consistency and control over the claims in an ad.
- Final URL expansion can change the intended customer journey, and it makes landing page relevance and attribution harder to trust.
- Expanded matching raises the question of which queries trigger ads and whether they fit the campaign's intent.
Google encourages advertisers to evaluate the full set of capabilities, and I understand the argument. A test with features switched off may not show everything the product can do. Still, I don't believe testing automation requires giving up controls that serve a legitimate business purpose. The right configuration depends on what an advertiser is trying to accomplish and the risks it's prepared to accept.
Automation can widen what an advertiser is eligible to serve against. Eligibility alone isn't a business outcome. Messaging, landing page experience, measurement and query relevance all still matter.
Why I'll keep testing
Search is changing, and AI Max will change with it. New search experiences, including ads in AI Mode, could shift its value. I plan to keep testing as it evolves.
Google has moved toward more automation before, and I'd rather understand a capability before it becomes a default. But I won't scale something because the platform is heading that way. I'll scale it when the incremental value shows up.
This reflects one set of tests, not a verdict on the product. Company details are generalized.