Initial tests conducted by Performics (Publicis) on campaigns from established advertisers reveal the need to optimize the tool from the outset.
Nearly a year after Google launched AI Max for Search, the first feedback is starting to come in. For the most experienced advertisers, this Google Ads feature can prove useful, provided it is managed from the outset. This is one of the conclusions reached by Nicolas Pestourie, Director of SEA Operations at Performics (Publicis), and Caroline Hajdukiewicz, Deputy CEO at Performics, after observing that AI Max easily generates incremental traffic but at the expense of cost per acquisition.
How does AI Max for Search work? It enables Google Ads, using predictive AI, to engage in “one-to-one” interactions with each user by understanding the meaning and intent behind every query and displaying the right ad copy and landing page that directs users to the advertiser’s site. All while taking into account what the advertiser is trying to promote. How? By leveraging Google’s access to the user’s search history and advertiser data (campaign history, website content, and even their CRM systems), among other signals. The holy grail, in short. This allows for expanding campaign reach without sacrificing relevance—all automated and fully autonomous. Sounds great on paper.
Performics conducted A/B tests to compare the performance delivered by AI Max for Search with the results obtained using traditional tactics on ongoing Google Ads campaigns for three advertisers: a major French retailer in the cultural sector, a provider of remote surveillance solutions, and an airline. During these tests, traffic gains—sometimes significant—were observed in all three cases: +10%, +16%, and +3%, respectively.
The quality of the traffic, however, leaves something to be desired for now: conversions are down 21% for the airline and 8% for the retailer, compared to traditional campaigns. “The quality of incremental traffic varies and needs to be verified through additional tests. In the cases of tourism and retail, we must also take into account the fact that the very large volume of search terms makes it difficult—even for AI—to identify the right queries on which to display the correct product URLs,” notes Nicolas Pestourie. “In this case, we need to guide the algorithm by feeding it the right page feeds or even by excluding URLs that are not relevant to the advertiser,” he adds. “In short, it’s better to guide the algorithm from the start rather than letting it cast a wide net,” summarizes Caroline Hajdukiewicz.
Only the remote monitoring solutions provider saw a 2% increase in conversions. This ultimately resulted in a 9% increase in its cost per acquisition (CPA). In the case of the retailer, with conversions down sharply compared to traditional search campaigns, return on ad spend (ROAS) was directly impacted, dropping by 18%. Finally, while the airline did not see a decline in its ROAS—which increased by 14% despite a 21% plunge in conversions—this was due to the fact that traffic growth was modest, at just 3%.
“These are very early tests; the results will undoubtedly improve as we implement optimizations with Google and as we receive feedback from other advertisers,” notes Caroline Hajdukiewicz. “Let’s not forget that AI Max for Search has allowed us to seek out additional traffic; it is the very nature of this feature to cast a wider net in search of new opportunities. It’s up to us to better guide the algorithm from the start,” she continues.
In fact, by expanding the semantic scope explored, AI Max for Search has sometimes allowed these experts to identify needs and themes that hadn’t been spotted at all until then. “Thanks to AI Max for Search, we’ve injected new relevant keyword themes into our campaigns, and that’s a positive development. It’s an extra fishing net,” confirms Nicolas Pestourie.
In AI Max for Search’s defense, it must be said that these tests set the bar very high, given that these core campaigns were already optimized using both broad keyword targeting and Performance Max—both of which are designed to significantly boost campaign performance. “We’re already seeing incremental conversions with broad targeting and PMax,” admits Nicolas Pestourie. This, too, makes the task even more complicated for the AI…
Hence the second lesson to be drawn from these tests: AI Max for Search can prove to be an interesting feature for smaller advertisers who are less experienced in managing various SEA tactics or who have less time to devote to them. The only requirement, however, is having a minimum amount of history on Google Ads to give the AI something to work with.
