AI Briefing: Index Exchange and Cognitiv integrate to use generative AI for programmatic curation

Two of the ad industry’s buzziest topics have generated yet another deal: generative AI and ad curation.

Last week, Cognitiv and Index Exchange became the latest companies to partner on another way to use deep learning algorithms and large language models for programmatic media. The deal lets Cognitiv’s ContextGPT product integrate with Index’s Marketplaces platform, which helps publishers use curation tools across demand-side platforms, data activation platforms and commerce media networks.

Rather than rely on keyword targeting and traditional audience segments, Cognitiv’s platform uses its deep learning and OpenAI’s LLMs to analyze millions of websites each day to help advertisers target audiences without relying on cookies. When Index sends an impression, Cognitiv provides real-time analysis to predict which ads will likely convert and then sends it back with an advertiser deal ID. Index then sends the impression to the advertiser’s demand-side platform before placing a bid.

“Advertising is the right message to the right person in the right place at the right time,” Cognitiv CEO Jeremy Fain told Digiday. “The audience is one of those things and context is another one of those things. But when they’re not combined together — especially with time and then maybe with message — it’s suboptimal. You’re showing an ad to the right person but in the wrong place or you’re showing an ad in the right place to the wrong person.”

To create custom audiences and content categories, advertisers use ContextGPT to input their product, desired audiences and the types of content they might want to buy ads across. Next, the platform uses LLMs to craft template prompts for generating audience segments based on a buyer’s media brief. A list of potential URLs is then generated, which buyers can sift through and choose to have either more or less of along with other filters for brand safety content sentiment. (The dashboard also shows inventory metrics like relevancy, daily impression estimates and unique URLs.)

Although contextual targeting and curation have been increasingly popular, Cognitiv boasts an ability to scrape and categorize new URLs every 15 minutes to provide a more real-time database for media buying. Fain said speed also helps advertisers target relevant ads and avoid unwanted placements while helping publishers increasing potential revenue per impression.

Using AI-powered curation could help attract differentiated demand from new advertisers and a wider range of ad budgets, said Index Exchange CEO Andrew Casale. AI curation that Cognitiv offers could help supply-side platforms like Index Exchange evolve from a commoditized infrastructure and drive innovation usually only associated with demand-side platforms, he said. Speed is important when considering user and contextual data; Casale noted Cognitiv can send signal back within just 5 milliseconds rather than the 150 milliseconds a DSP usually gets.

As AI curation becomes more integrated with SSPs, Casale thinks it could help drive faster universal adoption across the ad tech ecosystem since Index Exchange is plugged into about 100 different DSPs.

“We’re actually there at the beginning and we’re there at the end,” Casale said. “Which means if we enrich the bid request with a new signal like this, it immediately is available to any DSP. There is powerful innovation to unlock and we’re seeing that this is a great use case.”

This isn’t Index Exchange’s first integration with a company providing AI-powered curation. It’s also worked with Chalice AI, which is one of several companies providing SSPs with ways to use LLMs for contextual targeting. Meanwhile, Google announced its own curation tool earlier this month within Google Ad Manager — potentially bringing more validation and more competition to the space.

Generative AI has also been an investment area for CTV. Earlier this year, streaming giants and adtech vendors — including NBCUniversal, YouTube and Disney — announced new ways of creating AI-generated audience segments based on contextual understanding of onscreen content.

While curation itself isn’t new, the novel parts are “the promise of accuracy and the promise of competency,” said Simon Poulton, evp of innovation at performance agency Tinuiti. He thinks AI curation could help give advertisers and publishers more control that’s especially needed in the current era of black-box automation. However, he said the possibility of AI bias and potentially giving up some control remain challenging.

“Now we’re in this sort of maturity of space beyond the mad scientist era of, ‘Hey wouldn’t it be cool if we could do this’ to this is now a reality,” Poulton said. “There is still this big thing I do think about: the chasm-crossing moment of the degree of control we’re going to relinquish within these environments.”

To check for AI biases within the deep learning models and the LLMs, Cognitiv worked with inclusivity experts at the Linguistic Society of America. The company also worked with diverse groups internally to examine how the platform defined content and audience types based on gender, ethnicity and age. Another feature within Cognitiv’s platform analyzes sentiment of brand mentions on articles to make sure advertisers can avoid pages where they’re mentioned negatively.

“Every time you’re targeting any kind of ethnic minority, you want to make sure that you don’t show up on content that is stereotyping them or is talking about them in a negative sense,” said Mukul Yadav, head of product at Cognitiv.

Curation tools are key differentiators for publishers, said Forrester analyst Mo Allibhai. He also noted Index Exchange already has stringent publisher standards in place, which also helps lower the risk of AI hallucinations by LLMs. 

“Curating inventory typically involves three elements,” Allibhai said. “Deep, qualitative knowledge of the available inventory, meaningfully segmented audiences that buyers would look to activate, and connective tissue to those buyers. And it seems that Index Exchange is leveraging AI to scale efforts across all three fronts.”

Integrating AI-enabled ad curation across SSPs could help level the playing field of adtech, said Scott Messer, founder of Messer Media. However, he also cautioned publishers need to be wary of companies stealing their data with AI under the guise of contextual targeting and brand safety.

“The DSPs have too much power because they’ve been cookie monsters and they never needed to talk to a buyer or talk to a publisher before,” Messer said. “Anything that sort of weakens the grasp of a DSP in my opinion is generally beneficial for a publisher in reclaiming their power and ability to generate sales on their own.”

Prompts & Products — AI news and announcements

  • Jasper debuted new ways to give marketers more control over AI-created content.
  • Writer AI, and enterprise-focused generative AI startup with its own LLMs, raised another $200 million including from venture funds run by Salesforce, Adobe and IBM.
  • Haul, Amazon’s new Temu competitor, feature product images that seem to be AI-generated.
  • Google debuted a dedicated Gemini mobile iPhone app that lets people text and talk with the chatbot, create AI-generated images, access other Google apps. OpenAI also added new features for its ChatGPT app on Apple and Windows devices.
  • 7-Eleven partnered with AI startup Qsic to create localized audio ads across 5,000 stores to reach audiences at key times.
  • TikTok expanded availability of Symphony Creative Studio that lets advertisers create AI-generated videos.
  • Factiva, the Dow Jones-owned business research tool, debuted a new tool for AI-generated summaries powered by Google’s Gemini models. 
  • Coca-Cola debuted an AI-generated holiday ad to replace its original “Holidays are coming” ad. The new commercial air this holiday season across TV screens in the U.K.
  • Estee Lauder and OpenAI announced an expanded partnership, which includes a “GPT Lab” that has already made 240 custom GPTs to help employees analyze consumer surveys, develop new products, and market existing categories.

1s and 0s – AI research and stats

  • Lexicon Branding debuted new research to inform how companies should approach branding AI products.
  • Google added new Google Search support for metadata in AI-generated images that meet C2PA standards.
https://digiday.com/?p=560801

More in Media Buying

Media Buying Briefing: Looking at the implications of Omnicom’s IPG purchase, including if it doesn’t happen

What seems clear from the 10,000-foot point of view is that there are two general schools of thought around the acquisition, which would vault Omnicom to the No. 1 position.

AI Briefing: Autonomous browsing and shopping agents bring new opportunities and (bot) risks 

Google’s Project Mariner and other AI agents could change how publishers, advertisers, and e-commerce deal with bots — for better and for worse.

Data licensing lawsuit adds a legal wrinkle to Omnicom’s planned acquisition of IPG

There’s been a lot of speculation about the value of Acxiom to Omnicom’s acquisition of IPG, but an ongoing court case over the data warehouse adds another layer.