If someone says they’re about to drop a bomb, it could mean that it’s time to run for cover. But it could also mean they’re about to share a piece of juicy gossip. Context is, in fact, king.
Around two years ago, AI marketing company Cognitiv launched an AI-powered contextual targeting tool called ContextGPT with the aim of making campaign planning quicker and more efficient by matching advertisers with content that suits their brand goals.
On Wednesday, Cognitiv released the latest version of ContextGPT, which allows brands to provide more detailed and specific targeting parameters in order to optimize ad performance, Aaron Andalman, co-founder and chief science officer of Cognitiv, told AdExchanger.
Brand suitability used to rely solely on keywords, said Andalman, but Cognitiv’s technology understands content situationally based on how words are being used.
Out with the old
Although contextual targeting is far from new, using AI to identify placements allows Cognitiv to make real-time decisions and continually learn from newly ingested information, said Eric Danetz, the company’s president.
In previous versions of the tool, brands would provide an example of the type of content they wanted to advertise against – the first paragraph of a relevant article, for instance – and Cognitiv’s AI would offer up a list of URLs that matched the desired tone and other criteria.
The updated version features a chat-like interface where advertisers can describe their core audience using specific personality traits and behaviors, like young dog owners who don’t have kids but treat their pet as their child.
ContextGPT then uses its AI models to interpret the advertiser’s request and churn out contextually relevant suggestions of where to advertise.
The chat function provides a “more seamless integration with the media brief,” said Andrew Sisk, manager of performance planning at IPG-owned media buying agency Initiative, which has been testing the tool.
Now, Sisk said, fewer details get “lost in translation” during both the planning and activation process since advertisers can target specific demographic audiences, rather than just designing their ad strategy around preexisting content.
Context is the key
In addition to being able to chat with the tool, another new feature of ContextGPT is a brand’s ability to block specific types of content.
An electric vehicle company, for instance, could target Tesla-related content while opting out of stories about Elon Musk, and pet owners could ask to avoid any sad animal stories, like shelters posting about euthanasia, said Jennifer Newman, Cognitiv’s director of product management.
Cognitiv also upgraded what it calls ContextGPT’s “relevance engine,” which analyzes and filters potential content for context and brand suitability. It was built on a number of AI tools, including several OpenAI LLMs.
The new chat function is built on OpenAI.
Brands can still get a list of URLs to target against, but now it’s broken down into categories, such as by topic or strategy, to make it easier to filter for specific types of content.
For instance, brands can decide how particular they want to be about the content they’re advertising against. Say an advertiser only wants to reach that aforementioned audience of young dog parents who spoil their pets. They might select 95% on Cognitiv’s relevance scale, whereas buyers with looser specifications could opt for a lower number to expand their target audience.
Once an advertiser determines which strategy best fits their goals, Cognitiv creates a “dynamic deal ID,” Andalman said, which is appended in real time to bid requests through SSPs and then feeds into the client’s preferred DSP.
And the URL list isn’t static, he added; it’s being constantly updated. Meanwhile, a URL can be added to a deal within 15 minutes of appearing on the list.
“It’s the same buying mechanism as we had before,” Andalman said, just with more nuanced levers and outputs.