What 23 tests reveal about Google AI Max performance

We’ve tested Google AI Max over the past nine months, analyzing 23 individual tests across 16 already mature advertisers operating within a range of verticals. This article reveals what we did to maximize success with this campaign type.

Your experiments and observations may vary. If so, we’d welcome the debate.

This is intended to be just one voice among many in the conversation around AI Max. All the analyses we discuss are replicable within your own accounts, so you can ratify or dispute the findings based on your own data.

The ground rules for AI Max

Before launching an AI Max test, consider several factors. Two are particularly significant:

• Your campaigns should bid on a conversion action that’s meaningful for your business. Aim to get your conversion hygiene in as good a place as possible through tools like Enhanced Conversions and Google Tag Gateway. Value-based bidding is also ideal, although it’s not essential. Any automated targeting functionality can work.
• Your campaigns shouldn’t be budget-constrained. This advice is true in many situations, but it’s particularly relevant with AI Max. What’s the point of opening up your targeting if your budget prevents you from entering those auctions anyway? If your campaign is limited by budget, then either increase your daily budget headroom or set more conservative bid strategy targets.

With those prerequisites satisfied, we can now cover some of the juicier findings we’ve uncovered from our AI Max tests.

Learning 1: Go all in with AI Max

AI Max performs best when you enable all three core features simultaneously:

• Search term matching.
• Text customization.
• URL optimization.

Overall, we saw a 40% higher uplift in test success rates for campaigns that used all three features compared to those that opted in only to the baseline search term matching functionality.

Text customization drives stronger performance

Google has been pushing the text customization concept in various guises for a few years. However, earlier versions, like auto-applied recommendations, have had limited uptake. So, we were keen to finally assess the impact this would have.

Using the Added by segment in the assets report, you can compare how text customization performs compared to standard advertiser-provided assets.

We found that AI-edited assets delivered an improved return on ad spend (ROAS) and helped extract more value per impression. Put simply, clients were better off when text customization was activated than when it wasn’t.

This trend was consistent across both headline and description assets, even though we found that text customization modified headlines far more often than descriptions.

Text customization skews the auction in your favor

Strong performance is the ultimate objective for AI Max campaigns. But from a search geek’s perspective, the arguably more tantalizing result is that text customization demonstrably improved Quality Score.

We assessed historical Quality Scores for clients who activated text customization before and after the test launch. This analysis is valid because the Google Ads interface reports Quality Score only when the search query syntax exactly matches the keyword. This methodology provides a like-for-like comparison across a group of queries that were targeted both before and after switching on AI Max.

We saw a topline improvement in weighted Quality Score, from 6.8 to 7.3. This upward trend repeated across the three components of the Quality Score, with ad relevance showing the most notable uplift.

Logically, this shouldn’t be a surprise. After all, the premise of text customization is that Google shows the best possible ad to each individual user. Nonetheless, it’s satisfying to see this story unfold in our analysis.

At the same time, this finding is noteworthy because advertisers have generally been reluctant to use the full AI Max suite. Across all our test cases, only 50% used text customization, and even fewer (44%) enabled URL optimization.

Some brands will need to adhere to compliance guidelines that outright prohibit the use of these features. But our results suggest that if you have any wiggle room at all, you’d be well served by running a test with all three features.

Google is constantly rolling out additional guardrail features to clarify what is and isn’t off-limits from a brand messaging perspective. Marketers in more risk-averse organizations would be well-advised to keep a close eye on these releases.

Dig deeper: Google expands AI Max text guidelines globally

Learning 2: Take an account-wide approach with AI Max

This next suggestion might seem counterintuitive, but hear me out.

If you’re testing out AI Max for the first time, you might be better off enabling the feature across your entire account right from the start, rather than following a step-by-step approach. There are a few reasons for this.

Not all AI Max traffic is net-new

With AI Max enabled, you can target more queries and users than before. And of those queries, many will genuinely be net-new to your account.

However, it’s also common for queries that another campaign in your account once reached to get pulled into your AI Max campaign.

When we assessed performance at the campaign level, we saw an average +7% increase in conversion value, directly generated by queries the campaign had never targeted before.

When we zoomed out to an account-level view, however, only 46% of those queries were actually new to the account. The remaining 54% had previously been captured elsewhere in the account.

That still isn’t a bad result. An approximately 3% incremental uplift in conversion value, especially for accounts that were already running with a high broad match adoption, is great.

But this finding does have two key implications:

• If you care about your search term hygiene, enabling AI Max in only a subset of your campaigns could disrupt your search term-to-campaign funnel. Because brand inclusion lists are now exclusively available for AI Max-enabled campaigns, enabling AI Max account-wide can help you maintain a cleaner search term-to-campaign funneling system.
• Single campaign adoption muddies the water when assessing the success of your test. You care about net-new conversions, not reorganizing existing traffic within your account. When testing AI Max, make sure you assess the full account-wide impact.

---

How not to evaluate AI Max

Don’t rely on a cost per acquisition (CPA) by match type analysis to assess AI Max’s efficacy. This approach reveals attribution data within your campaign. But what you really want to know is whether AI Max has improved your overall ability to generate returns at an incremental investment that you’re comfortable with.

There are examples of advertisers trialing AI Max and achieving account-wide efficiency improvements. But you should identify those cases by reflecting on macro, account-wide performance — not by looking at your match type CPAs.

Why you should monitor campaign types

Consider how AI Max interacts with your other campaign types and targeting methods. Let’s call out one particularly glaring example: Dynamic Search Ads (DSA). In our own analysis, every successful AI Max test occurred in an account with low-to-no adoption of DSA campaigns.

This is understandable. Almost every single capability of DSA campaigns is now available in AI Max. So, it shouldn’t be surprising that having both campaign types running in parallel doesn’t improve performance.

It’s plausible that we may not be that far away from Google announcing another round of campaign streamlining initiatives, similar to those for Smart Shopping and Discovery campaigns in previous years. But until then, it’s on marketers to put some thought into the role you intend each campaign type to play within your overall account plan.

Dig deeper: AI Max in action: What early case studies and a new analysis script reveal

Learning 3: Think beyond AI Max

If you’re already comfortable with AI Max and you’re ready to push onto the next step, there’s a wealth of new testing opportunities to think about.

Search Bidding Exploration (SBE) was and still is the first major user-facing change to Google’s bidding technology in the last five years. Yet there’s been remarkably little industry chatter so far about this feature. SBE feels like a natural partner for AI Max, given that both tools are designed to reach incremental and previously inaccessible customers.

AI Max also gives you the chance to evolve your thinking around account structure. In an AI Max world, the optimal balance between segmentation and consolidation may lie elsewhere than before.

We’re already starting to see some green shoots of successful hyper-consolidation approaches. But it’s still too early to decisively comment one way or another.

Dig deeper: AI Max increases revenue 13% but drives higher CPA: Study

Putting AI Max to the test in your own account

It’s an intriguing time to be working in paid search, and AI Max has already sparked significant debate and experimentation within the industry. If you’re a later adopter or if you’re looking to improve on a previously unsuccessful foray into AI Max, then consider the following:

• Implement key ground rules: Ensure that you have objective-oriented bid strategies in place, powered by strong conversion hygiene. Remove campaign budget constraints once and for all.
• Adopt an all-in approach: Text customization and URL expansion may not be as popular as search term matching. But we’ve observed that using the full package can actually improve the likelihood of success — by up to 40% in our experiments.
• Prioritize an account-wide impact: Consider the interplays between AI Max, your regular keyword campaigns, and DSA. It might be that an AI Max everywhere approach is preferable. When judging results, look beyond campaign-level tests where possible, and block out the CPA-by-match-type brigade.
• Get creative: Think about the more innovative ways you can integrate AI Max with other facets of your account.