DIVE Blog - AI Optimization

One of the unique aspects of Google Performance Max (PMax) is that Google, the seller, isn’t just distributing the ads but is also actively optimizing them. This raises an important question: Can we truly rely on the same party that sells the advertising space to also optimize our investment in it?

When a company is both the vendor and the optimizer, there's an inherent risk of bias. It’s in Google’s interest to maximize ad spend across its inventory, and while AI promises efficiency and performance, there’s a fine line between optimizing for business goals and simply maximizing revenue for the platform itself. This can lead to campaigns being skewed towards metrics that benefit the seller—such as brand protection or broader reach—rather than pure conversion-driven results.

Google markets PMax in three steps:

• 1. Find more converting customers with a single, consolidated campaign.

• 2. Achieve better results with less effort.

• 3. Leverage the power of AI.

While the allure of AI (bullet #3) is real, the promises of delivering better results with less effort (bullet #2) and finding more converting customers (bullet #1) are not always aligned with the advertiser's goals. In fact, many advertisers report less-than-ideal outcomes when leaving everything in the hands of Google's AI.

At Deep Dive Engine, we strongly recommend caution. We’ve found that relying solely on Google’s reporting can often present a rosier picture than reality. In fact, 9 out of 10 Performance Max campaigns we've audited revealed potential cost savings, suggesting inefficiencies that may not be visible in the standard reports provided by Google.

Our advice? Anyone using Google Performance Max should enlist an external, unbiased auditor to assess their campaign performance independently. Having a third party review the data ensures that your interests as an advertiser are protected, and that you're not overspending due to hidden inefficiencies or biased optimizations.

What are your thoughts? Have you experienced similar concerns with Google Performance Max? Please don't hesitate to get in contact.