Nicholas Thompson, chief executive officer of The Atlantic, kicked off the Thursday conference sessions of the INMA World Congress of News Media in New York with a discussion on the biggest questions AI poses for the media.

Direct audience relationships in the AI era

Thompson addressed the existential threat that faces the Web as a result of AI. Given the dependence of news outlets on search engines to pull visitors into their Web sites, media companies are being forced to reassess their business models as consumers switch from searching for Web sites on Google for information to simply asking ChatGPT.

“This is the most important question that we’ve been dealing with on the business side at The Atlantic,” Thompson said. “In a world where the Web is fundamentally different, how do we develop direct relationships with our readers? How can we make our business resilient in a world where the fundamental thing that we use to drive our business doesn’t exist?”

Thompson cautioned against the inclination some media companies may have to try to preserve as many of their online visitors coming from search engines as possible. He noted that the overall traffic on the Web is declining for the first time in two decades — and that trend does not show any sign of slowing down.

“We don’t need to keep all of them. We just need to keep some of them,” Thompson said of readers from search engines. “I don’t mind if we lose 90% of our Google traffic but we keep the 10% who have some propensity to subscribe or become loyal readers, that’s fine.”

AI as a journalist’s assistant, not a replacement

While Thompson acknowledged the great value that AI can bring, he lamented the tendency by companies to use AI to emulate humans, rather than utilising the technology as a tool to assist and support humans.

“Instead of using this miraculous technology to build tools, they centered all this money in trying to build machines that are as much like us as possible,” Thompson said. “You can see it in the first time that ChatGPT came out. It didn’t have to type like a human, but they thought it’d be better if it typed like a human.”

Thompson presented his preferred use case for AI as a journalist: inputting large volumes of notes, transcripts, and data into an AI solution and prompting it to check for mistakes or gleam the most important insights and quotes.

“Please read this entire manuscript and tell me whether there are any moments where I use the same word within 20 words of each other,” Thompson said as an example of how he would prompt an AI model. “Use it as a highly intelligent editor who is on call 24 hours a day.”

Opportunities and risks

From a business perspective, Thompson discussed the prospect of an AI agent visiting a news Web site rather than a human. How can media companies optimise their business models to monetise this emerging non-traditional consumer?

“If an agent is visiting my site, I’m not going to show them travel ads with images of Bermuda. I’m gonna show them books to buy,” Thompson said. “What we need to be able to do is distinguish the two so that I know if it’s an agent or a human. Then I can decide maybe I want to block you, or maybe not because I think I can sell you a book or maybe I have a syndication deal with the agent’s parent company.”

Although AI unlocks significant opportunities, Thompson shared his concern about the growing challenge of cognitive offloading. Similar to how using a calculator can degrade a student’s own ability to do math, Thompson worries about how an over-dependence on AI can erode users’ critical thinking skills.

“Have you ever been in a meeting where somebody asks something complicated, and your instinct is to go ‘I can’t wait to get back to AI?’ That’s what’s scary,” Thompson said. “It makes you smarter when you’re with it and dumber when you’re not.”

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .