Testing Google AI Max: Incremental Growth or Incremental Risk?

Artificial intelligence is rapidly reshaping paid media. And Google’s rollout of AI Max is certainly no exception.

Built within existing search campaigns as a one-click “upgrade” feature, AI Max uses artificial intelligence to discover untapped queries through “keywordless” matching by recognizing additional intent signals beyond the limits of traditional keyword targeting, while also allowing real-time text customizations and final URL expansion capabilities.

It promises stronger performance and increased conversions through tailored ad experiences and delivery on high-performing queries that may otherwise be missed.

In theory, it sounds like the next evolution of search.

But in practice? We’re not convinced.

We ran a controlled test to evaluate AI Max’s ability to drive conversion growth through expanded query reach. Here’s what we learned and why we believe marketers should proceed with caution.

The Test: 

We enabled AI Max within a non-branded responsive search ad campaign optimized for conversions with the following guardrails applied:  

• Brand exclusions in place 
• Text/URL expansion controls applied 
• Keyword targeting (phrase match) left unchanged to run alongside AI Max 

Our objective was simple: determine whether AI Max could drive incremental, efficient conversion lift beyond what the campaign was receiving through keyword targeting alone.

Specifically, we wanted to evaluate its ability to use additional intent signals—such as user behavior (e.g., past searches), contextual cues (e.g., location), and other patterns—to identify and capture high-value demand (searches from people likely to convert) that keywords may miss. 

The Results:

1. Efficiency Declined

Instead of incremental lift, period over period we saw: 

• Conversion rate drop from 4.7% to 3.6% 
• Total conversions decline by 10.4% 
• Conversion costs increase by 4% 
• Slightly improved cost per click (CPC), but that’s not the metric that mattered. Efficiency eroded.  

Lower CPCs don’t compensate for weaker intent.

2. Irrelevant Query Expansion Became a Problem

With AI Max enabled, our test campaign began serving ads on broader and more vague search queries where user intent often misaligned with the campaign’s intent and the brand’s positioning.  

For example, imagine a campaign for a dentist targeting searches like “teeth whitening near me.” With AI Max enabled, ads might appear on broader queries such as “tooth cavity prevention” or “dental insurance.” These searches stem from very different intent than someone actively looking to book a cosmetic dental service and might not even relate to any services the dentist provides.  

While hypothetical, this example mimics similar results observed in our test where compared to traditional keyword match, AI Max underperformed across the board.  

Even with utilizing a maximize conversion objective, Google’s algorithm appeared to prioritize reach over precision, expanding into terms that lacked clear intent. The “high-value intent signals” often missed the mark, matching ads to users seeking entirely different solutions from what the ads sought to provide, weakening alignment between user intent and ad delivery. 

Furthermore, after nearly four weeks, the campaign showed little ability to get “smarter” over time, continuing to serve ads on the same low-intent queries that consistently failed to generate conversions. 

This is a critical distinction: Volume does not equal incremental value.

3. It’s More than Traditional Keyword Search “Backup”  

Google positions AI Max as a supplemental layer to traditional RSA query matching, designed to capture incremental opportunities missed by keywords. In theory, it should operate in the background, expanding coverage only when conversion probability is higher. 

With this designed intent, it should: 

• Be selective 
• Augment, not interfere 
• Avoid cannibalizing proven traffic from keyword matches 

But that’s not what happened. Instead, AI Max accounted for 20% of identifiable search term spend during the test.  

Yet it delivered:  

• Lower conversion rate  

• Higher cost per conversion 
• Lower overall efficiency  

When a system consumes that much of the budget, it must justify its presence.  

And in this case, it didn’t.

4. It Didn’t Unlock New Scale. It Constrained It.  

Perhaps the most surprising outcome of the test wasn’t the decline in efficiency, but the loss of scale. 

Historically, this campaign had no difficulty utilizing its allocated budget. Spend was stable. Volume was predictable. 

After enabling AI Max, that stability changed. 

Delivery became more volatile. The campaign began struggling to deploy the same daily spend, leaving dollars on the table despite no change in budget or constraints. 

That contradiction matters. 

If AI Max were truly unlocking incremental opportunities as a supplemental layer alongside our keywords, budget utilization should have become easier if anything. 

But instead, we saw the opposite. 

5. Questionable Incrementality 

Lastly, we also observed some overlap between AI Max and traditional search terms.  

This raises an important question: Are all AI Max matches conversions truly incremental?  

If AI Max competes with or duplicates keyword coverage, then it steals delivery from covered keywords rather than expands it.  

Without clean incrementality, the value proposition weakens significantly.  

The Takeaways: 

The promise of AI Max is compelling: capture new intent, boost conversions, reduce manual management.  

In reality, it isn’t accurate or mature enough for reliable growth. 

Based on our test:  

• It struggled to match high-intent relevance.  
• It diluted efficiency.  
• It failed to learn or was simply too slow to adapt. 
• It consumed meaningful budget.  
• It did not demonstrate clear incremental lift.  

That doesn’t mean it will never work. It means it’s not ready to be trusted blindly.  

If you’re considering testing AI Max, we’d recommend taking a cautious and controlled approach:  

• Run controlled experiments with select campaigns, not full migrations.  
• Monitor search term quality closely.  
• Measure incrementality, not just reported conversions.  
• Watch conversion rate and CTR, not just CPC.  
• Cap spend exposure until performance stabilizes.  

Our POV: 

Automation isn’t inherently bad. In fact, it can be powerful. But automation without precision, transparency, and human oversite naturally introduces risk.  

As platforms race to launch new AI optimizations, marketers are more frequently asked to place their faith in black-box systems and trade their control for promised scale and efficiency. That trade only works when performance justifies it.  

And that trust needs to be earned, not assumed. 

In our test, AI Max didn’t earn that trust. If anything, it raised a broader concern about where search is heading. 

At its core, Search is the most intent-driven channel in marketing. It works because ads are triggered by clear user intent (usually through keyword targeting), creating a direct link between what users are looking for and the solutions a brand provides. 

AI Max begins to blur that relationship. 

Instead of matching ads strictly to a user’s query, it can serve ads based on predicted interest signals. In other words, the system may prioritize what it thinks a user might want rather than what they are explicitly searching for in that moment. 

When that happens, the connection between user intent and ad delivery starts to weaken, making search less relevant (and more frustrating) for users and less efficient for advertisers. 

If this trend continues, search could begin behaving less like demand capture and more like demand prediction, effectively pushing it higher up the marketing funnel. 

That’s a meaningful shift and one marketers should evaluate carefully before handing over the keys.