“As you know, contextual ads have been around for ages,” Chow said during his talk. “But you also know they’re typically limited to a pretty shallow understanding of actual context.”
Chow said they used machine learning to overcome the shortcomings that have hindered “contextual” advertising in the past. The search giant worked with The Guardian’s creative development team to place ads in the recipes section of the publisher’s website. They had to teach machines to identify meals in the photos of recipes as either “sweet” or “savory.” Then the ads for Google Home Mini, the voice-activated speaker, would change depending on the context. The ads were “built dynamically, in real time as the site loads,” Chow said.
Google as a company hasn’t always been so “pro contextual ad.” In fact, earlier this month the company issued a defense of more highly personalized ads, saying that they generate greater value for publishers. Google has warned against eliminating internet tracking altogether, since data and targeted ads are the backbone of its advertising business. Google’s ad technology powers advertising on millions of websites.
In a report this month, Google said it studied ads that used cookies, which are the files that websites use to track consumer behavior online, and ads without cookies. It claimed that ads without cookies generated half as much money for publishers.