© 2024 Interlochen
CLASSICAL IPR | 88.7 FM Interlochen | 94.7 FM Traverse City | 88.5 FM Mackinaw City IPR NEWS | 91.5 FM Traverse City | 90.1 FM Harbor Springs/Petoskey | 89.7 FM Manistee/Ludington
Play Live Radio
Next Up:
0:00
0:00
0:00 0:00
Available On Air Stations

AI-generated articles are permeating major news publications

AILSA CHANG, HOST:

So who actually writes the news that you consume every day? Well, here at NPR, it's reporters, producers, editors, literally hundreds of human beings in our newsroom. And until recently, it was pretty safe to assume that real people were behind the bylines and the articles that you read. But last year, an investigation by the publication Futurism found that Sports Illustrated had been publishing AI-written pieces by make-believe journalists. It was a scandal that rocked the publication, and, shortly after, the magazine's publisher fired the CEO.

Reporter Maggie Harrison Dupre broke that story, and she has kept digging since, finding that AI-written content has made its way into many more publications, including the LA Times, the Miami Herald and Us Weekly, to just name a few. Maggie Harrison Dupre joins us now. Welcome.

MAGGIE HARRISON DUPRE: Thank you. Thank you for having me.

CHANG: Thank you for being with us. So you write about one major company behind this AI content. It's called AdVon. Tell us a little bit about what they do.

HARRISON DUPRE: So AdVon is a - what we call a third-party media contractor. Say you, for example, want to buy a vacuum. Your vacuum broke. You want a new one, but you want the best one. So you go online. You go to Google. You search, you know, best vacuums. You get a bunch of links. There is a big chance that AdVon has actually been the one who's created an article about best vacuums that's ended up in a variety of large news publishers in the U.S.

What we discovered in our reporting was that AdVon was, in many cases, bylining or attributing authorship to fake writers, in many cases with fake, AI-generated faces. And they also, we discovered in our reporting, were using AI to generate a great deal of this content as well.

CHANG: Right. And some of these AI-written reviews have some pretty strange mistakes in them. Like, you found some reviews they generated for weightlifting belts. Tell us about that (laughter).

HARRISON DUPRE: Yeah. So the weightlifting belts were a great example. You know, you want to go to the gym. You want to protect your back. They are very different from, you know, a traditional, hold-up-your-pants kind of belt.

CHANG: Yes. Yes.

HARRISON DUPRE: But we found a string of these reviews published in a local regional newspaper owned by the McClatchy media network, which owns a lot of newspapers throughout the U.S. - published reviews that, you know, the products that they listed in the buying guides were weightlifting belts. But about halfway through, it switched into talking, you know, inexplicably about traditional belts, even suggesting, you know, you can buy a great belt from Hermes, which is true...

CHANG: (Laughter).

HARRISON DUPRE: ...But it's not helpful in the context of this article.

CHANG: Not sporting that at the gym while you're lifting weights.

HARRISON DUPRE: Yes.

CHANG: I'm curious, how much reach does AdVon have? Like, just how widespread is this practice of buying its content?

HARRISON DUPRE: AdVon's spread is quite large. It has a large number of publishers in the U.S. It ranges from big regional newspapers - for example, the LA Times or the Miami Herald - to a lot of smaller regional publishers around the U.S. - The Raleigh News & Observer, for example - you know, a midsize city paper - to just a lot of small newspapers. USA Today is another big AdVon contractor, as well as, of course, Sports Illustrated previously. Yeah, its reach is quite vast.

CHANG: So when you contacted AdVon, what did they say to you about all of this?

HARRISON DUPRE: AdVon's story evolved over time. When we were first investigating Sports Illustrated, AdVon, you know, denied any use of AI in editorial content in full. They said, we even use anti-AI software still in all of our editorial content, so firmly no AI.

And then, soon after that, we found an AI training video in which an AdVon manager is seen using a, like, specifically developed AI system to churn out content that was exactly like what we were finding at Sports Illustrated and at the Miami Herald and at all of these different publications that we tracked AdVon content to. And at that point, AdVon's story did change. They admitted - they said, yes, we are using AI, but only when publishers say it's OK. Our reporting said, you know, perhaps this was not the case.

CHANG: Yeah, tell us about that. Like, what about the news publications running this material? How did they defend it, or did they change their minds once they found what was going on?

HARRISON DUPRE: So I think McClatchy is a good example of that. You know, we'd been back and forth with McClatchy for a while. They said that AdVon had assured them that there was no AI used. When we presented them with this video that we'd found, they did a review, and they chose to remove all of the content from their websites. Outside Inc., which owns Yoga Journal and a few other publications - they were contracted with AdVon but said they had no knowledge of AI if it was used. The LA Times said the same thing.

So when we brought this full picture of our reporting to AdVon and we said, we have found this training video. We've spoken to publishers, many of whom have - you know, again, like McClatchy - they deleted the content. AdVon stopped responding to our questions after we kind of gave them the full picture.

CHANG: Is there a way that we, as consumers, can spot an AI-generated piece? Is there something that's a telltale sign that, oh, this is totally AI? What would you suggest?

HARRISON DUPRE: I think, in general, you should always be looking at the byline no matter what, but maybe zoom in a little. See if something feels a little uncanny about it. The text might be super repetitive, stilted and weird. You know, I've spent a lot of time, of course, in these articles. They generally feel like an alien came to Earth and had never used a vacuum, does not even understand what a vacuum is...

CHANG: (Laughter).

HARRISON DUPRE: ...But somehow has access to every single piece of information about a vacuum that's ever been created and bottles that into a review - like, correct information but a lack of understanding or expertise.

CHANG: Yeah.

HARRISON DUPRE: I think that's what's really...

CHANG: Interesting.

HARRISON DUPRE: ...Salient in the content that we reviewed. And so I think, you know, if you're shopping for a vacuum, maybe just try to find one with a really good return policy.

CHANG: (Laughter) Well, this last question I ask with some amount of self-interest because this is about my profession. You know, like, this is a time when a lot of media organizations are financially struggling. I'm curious if reporting out this story has made you think differently about the future of journalism and how replaceable we humans are in this business.

HARRISON DUPRE: Spending this much time in this kind of content, in these articles, has certainly, you know, instilled a new sense of paranoia. At the same time, I do want to stress that this isn't journalism. Like, these articles are not good. And this is not journalism...

CHANG: (Laughter) Yea. Merit still wins out.

HARRISON DUPRE: Yeah.

CHANG: OK.

HARRISON DUPRE: Yeah. I think AI is just going to be - personally, I would argue that it's going to continue to become embedded into our lives in different ways. But in terms of reporting, you know, AI is not a person. It doesn't have personal experiences to draw on. And so I think in terms of news and reporting and really being creative, I would personally argue that that's still a deeply human endeavor. And to create good news, there needs to be a lot of humanity infused into it.

CHANG: Yeah, and lived experiences.

HARRISON DUPRE: Yes, exactly.

CHANG: That is Futurism reporter Maggie Harrison Dupre. Her article is titled, "Meet AdVon, The AI-Powered Content Monster Infecting The Media Industry." Thank you so much.

HARRISON DUPRE: Thank you for having me.

MARY LOUISE KELLY, HOST:

And we reached out to AdVon for comment on their use of AI across publications. A spokesperson sent a statement, which reads in part, quote, "we are committed to working closely with our publishing partners to ensure that their optional use of our AI solutions meets their content standards."

(SOUNDBITE OF MOTOR CITY DRUM ENSEMBLE'S "DETRIOT") Transcript provided by NPR, Copyright NPR.

NPR transcripts are created on a rush deadline by an NPR contractor. This text may not be in its final form and may be updated or revised in the future. Accuracy and availability may vary. The authoritative record of NPR’s programming is the audio record.

Kathryn Fink
Kathryn Fink is a producer with NPR's All Things Considered.
Christopher Intagliata is an editor at All Things Considered, where he writes news and edits interviews with politicians, musicians, restaurant owners, scientists and many of the other voices heard on the air.
Ailsa Chang is an award-winning journalist who hosts All Things Considered along with Ari Shapiro, Audie Cornish, and Mary Louise Kelly. She landed in public radio after practicing law for a few years.