AI could help improve ad targeting in connected TV, according to findings from a newly formed research group designed to study the utility of the emerging technology for TV advertisers. 

As advertisers battle increasingly fragmented media and ever-fluctuating numbers of ad-supported streaming subscribers, expanding ways to hyper-target TV audiences is critical. To many, successful targeting means audience data and threading a delicate needle of demographic and attribute parameters, but considering the storytelling surrounding a campaign can have just as large of an impact as reaching a single in-market person.

The Alliance for Video-Level Contextual Advertising released its inaugural research into the use of emerging technology for categorizing and tagging TV content for brands to better gel with the shows they’re advertising in.

Formed this year between ad tech and data companies Iris.tv, Beachfront Media, Pixability and Silverpush, the AVCA’s first research project found ad targeting by content attributes generated by AI can help advertisers better achieve their campaign goals. For example, ads served by AI-generated content metadata were watched four times as often as those targeted by broad demographic or by a publisher’s own data, and for the ads watched, AI contextual targeting commanded 15% more ad attention than the others.

To do this, AI combs through video content to extract data based on the story, characters and images on screen to create targeting segments for advertisers, which serve to both align brands emotionally to the storytelling viewers have tuned in for as well as to increase brand safety in streaming. And for brands that may not have access to robust stores of first-party data for hyper-specific audience targeting, AI-generated contextual targeting can help reach specific audiences by content terms instead, said Rohan Castelino, chief marketing officer, Iris.tv, and principal member of AVCA.

An example from the study, which placed eye-tracking glasses on members of households watching a TV-MA sitcom, found a negative response from viewers who were served a pharmaceutical ad for a serious condition that misaligned with the comedic scene that preceded it. Rather than block all comedy from the pharma brand’s campaign, it can utilize AI to programmatically serve different creative variants by the type of content so that the ad is most relevant to the viewer seeking out a comedy versus a serious drama.

And for those that saw the ad and aren’t in-market for that type of medication, viewers who saw ads served by AI-generated content tagging showed two times higher unaided awareness and are more likely to recommend that product, Castelino said.

Brand safety

AI-generated contextual tagging can also better help in creating more consistent brand safety parameters across partners. Castelino said that while streaming publishers share broad data on content by genre and rating, a modern PG film and one from the ’80s are vastly different. AI removes the subjectivity of content labeling and creates fidelity across content and brand suitability labels across platforms.

Recent news: X pushes vertical videos to attract brands

Castelino noted that while the AVCA’s research focused on traditional ad breaks, AI-powered contextual targeting can also be applied to other streaming native formats, such as pause or frame ads, plus developing formats such as shoppable ads.

“In addition to being able to find people wherever they are, it’s incredibly important that [brands] understand where their ads are running, because that’s going to have a direct impact [for business outcomes] and a long-term impact on brand perception,” said Castelino. “Bad data in means bad outcomes out, so the level of precision that AI offers is necessary because the different definitions for comedy, it’s a chasm … especially if you’re trying to buy programmatically and make a decision under 100 milliseconds. Content relevance is important, and AI is a necessary condition.”