Tl;Dr
We ran a three-month A/B test comparing Google’s new AI Max feature against our existing exact and phrase match structure. The results were pretty disappointing, leads dropped by 88% and CPA more than tripled. While AI Max shows promise as the technology develops, our advice for now is to test it carefully, with strict spend controls and clear success metrics in place.
What is AI Max?
In May 2025, Google Ads introduced AI Max, a new feature promising to change the way we think about search advertising.
Unlike traditional match types, AI Max uses keywordless technology to handle both search term matching and asset optimisation automatically. Rather than relying on predefined keywords, it learns from your campaigns and tailors your creative message to each individual customer, with the goal of driving more value for your business.
On paper, it sounds like a step forward. But does it actually deliver?
Test Overview
To give AI Max a fair shot, we ran a structured A/B test, pitting it directly against our existing exact and phrase match terms within the same campaign.
Rather than testing on a random campaign, we chose our key lead-driving campaign to get the most meaningful results. The test ran over three months, giving both sides enough time to learn, stabilise, and perform.
The Results
The results, while eye-opening, painted a clear picture. Despite a relatively stable CPC, with AI Max keywords costing just £0.10 more per click, the overall performance told a very different story.
Leads fell by 88% during the test period and CPA climbed by 272% vs our current structure, which was difficult to ignore.
The drop in CTR was perhaps the most telling signal. With CPC remaining largely consistent, the decline in engagement suggests that AI Max’s search term matching simply wasn’t connecting with the right audience at the right moment, serving impressions, but failing to inspire action.
Summary
Is it too early to write AI Max off entirely? The technology is still in its infancy, only just leaving Beta in April 2026, and like many of Google’s AI-driven features it may well improve as it matures.
Our recommendation is straightforward: if you’re considering testing AI Max, do so cautiously, with clearly defined spend limits and a robust measurement framework in place.