How Forbes is using ChatGPT referral data to create audience cohorts

“Semrush and Similarweb have introduced pretty good tools that give you information like, what are the prompts that are leading to your content, how much traffic are you getting from them,” said Forbes chief innovation officer Nina Gould on stage at the Digiday Publishing Summit Europe in Lisbon, Portugal, on Oct. 29. She noted that this data isn’t available from all AI platforms, specifically citing Google’s AI Mode as a black box because the referral data from Google’s ChatGPT-style search experience is indistinguishable from traditional search referral data.

Google’s AI Mode aside, Forbes is able to use the information provided by Semrush and Similarweb to create audience cohorts from the people coming to its site from AI platforms. 

“We create cohorts, meaning we understand the types of searches and the types of people that resonate. We don’t target them or track them currently when they come back. So it isn’t a [personally identifiable information issue] at this point. It’s just data points for your audience and data points for your editorial team,” said Gould.

Those data points show what types of content from Forbes resonates most for audiences on AI platforms to the point that those people are willing to click a link from ChatGPT or Perplexity to visit the publisher’s site. That content is primarily what Gould described as Forbes’ “core content”: “our lists, wealth reporting, our entrepreneurial stuff, our investing stuff.”

Beyond the content categories that are of most interest to this AI-driven audience, Forbes has also been able to glean behavioral differences. 

“They expect a really tailored experience. And I think, in a lot of cases, they expect some brevity. They expect to be able to get more information more quickly and more completely in one place,” said Gould.

To that end, Forbes has been updating its site to make its web product more useful for its audience. 

For example, Forbes has updated its “More For You” article recirculation unit that appears on article pages. That unit used to be purely contextual. “Now we’ve switched that over to the user-based recommendations as well. And I think that was like a 300% lift in [clickthrough rate] there from contextual,” Gould said.

That article page personalization actually stems from a redesign of the Forbes home page to be more personalized as well as modular, so that home page elements can be inserted into article pages. After updating the home page to make its news feed more personalized, Forbes saw a 45% increase in clickthrough rate, Gould said.

Publishers like Forbes may need to seek out other ways of updating their sites for the AI era, though. 

While ChatGPT et al. may be citing articles and sending traffic, publishers need to make sure the AI platforms are accurately citing their reporting. According to research conducted by the BBC and European Broadcasting Union, more than a third of adults in the U.K. believe the publisher is responsible when an AI platform provides erroneous information based on the publisher’s content. 

As a result, publishers would benefit from being able to emphasize the accurate information to the AI platforms scraping their sites. One method would be updating the HTML schema behind web pages for publishers to create a standard way to signal an article’s contents to an AI platform, similar to how the schema exists for publishers to provide content signals to platforms like Google and Meta. 

Unfortunately, no such standard exists or is even visible in the offing. But it will need to emerge sooner than later. The AI era for publishers has materialized much more quickly than seemingly anyone anticipated.

“That’s the thing that’s challenging for all of us is our business models are shifting very, very quickly and then everything we have to do in order to be competitive and visible, honestly, in this new landscape is shifting,” said Gould. “So we are learning as we go.”