The problem with AI-powered campaigns

Over the last couple of years, we’ve seen Google’s Performance Max and Meta’s Advantage+ drive increasingly strong results as the platforms learn more about consumers, and put the right ad in front of the right person, at the right time. However, there has always been some skepticism around the conflict of interest: the platform, which owns the inventory, deciding when to place which ad to which user. This gives a lot of control to the entity ultimately sending the invoices.

The last two months have seen gradual declines in Google’s Pmax performance. We’ve noticed it across the board, and online discussions have been voicing similar findings. We have clients whose Pmax campaigns were being outperformed by customised, manual Shopping campaigns, or traditional Search campaigns.

So what’s changed?

According to Google, there were some bidding and budget updates in mid-August which changes how Google handles campaigns which are “limited by budget”. Instead of bidding on the highest intent auctions to save money, the algorithm became more rigid and forced campaigns into lower-converting auctions.

But what we’re seeing here is the control the skeptics were worried about being exploited. We have Pmax campaigns receiving a higher proportion of spend, but high quality inventory on search results (Shopping and Search) decreasing with the shift to LLMs. This forces AI-powered campaigns such as Pmax to use YouTube, Display Network, and other upper-mid funnel channels, which have far higher inventory levels, to spend budgets. High-performing, lower funnel placements that target a user when they are in market are deprioritised, and performance drops as a result.

This same situation can happen in Advantage+ campaigns, when a high-performing audience or placement becomes overly competitive. Meta has the flexibility to show ads to lower-performing audiences or across secondary placements.

It’s critical that brands include a combination of AI-powered campaigns with traditional ones. What you know about your brand will always exceed what a platform knows. Things like targeting by demographic, being specific with placements, and manually optimising campaigns will be what allows brands to drive incrementally better results over a long period of time, rather than relying on the short term sugar hit of automated campaigns.