Big-name news outlets – including the BBC, Newsweek, and Fortune – have been duped into using dozens of AI-generated “subject matter experts” as reliable sources in hundreds of published news articles.
That’s according to a new deep dive investigation published Monday by veteran tech and science journalist Rob Waugh of the Press Gazette, a UK media trade publication.
Waugh found dozens of instances where "digitally created" professionals have been quoted as experts in articles written by human journalists and published by major news and media outlets – including the BBC, The Guardian, Newsweek, Fortune, The Sun, Vogue, Marie Claire, Good Housekeeping, Women’s Health, Medium, Pop Sugar, and the list goes on.
Posting about the findings on LinkedIn, Waugh declared the entire fact-finding mission to be “quite a rabbit hole.”
“When you use media request services, you might assume you are speaking to a real, live human being… Sadly, there is now a good chance you are talking to an AI-enhanced fake,” Waugh writes.
“These is [sic] really worrying and frightening in equal measure... Not least because it plays into the whole conspiracist mantra of 'you can't believe/trust the MSN' [mainstream media],” commented David Wilson, PR Director at NetEase Games.
Waugh believes it's all about algorithms and the coveted SEO (Search Engine Optimization), which determines how high a website ranks in a browser's search results depending on the specific keywords or phrases users search for.
“The benefit for those who reply to these journalist requests is a name-check for the brand they are associated with and sometimes even a link, which is valuable for SEO,” he said.
Experts in, well, everything
In one instance, Waugh noted a so-called “psychologist” with a degree from Oxford, Barbara Santini, had been “featured literally hundreds of times in the British media, talking about everything from the psychological benefits of sleeping with your dog to 'daycations.’”
Attempting to follow up and confirm her existence, Waugh said the alleged mental health professional, “refused to confirm her qualifications - or speak on the phone." In fact, the only way to communicate with the UK's 'most featured' therapist was through WhatsApp.
Another purported expert commentator, “Rebecca Leigh” of Academized – a real company that assists students in writing papers – was found to be an AI fake (freely admitted by the company itself). Leigh offers expert commentary on everything from the carbon footprint of avocados to employee benefits to online music streaming, according to Waugh.
When asked by Waugh to validate herself as a human entity, "Rebecca also stopped communicating,” he said.
“So not only is AI giving us fake answers, we're being served fake experts in non-AI spaces? I was all set to be outraged, then I realized this is actually only a problem of scale. Fake experts have been around at least as long as snake oil,” creative director Ben Levy also posted on the thread.
Waugh explained that Leigh, who has already been quoted extensively in the media (Fortune, Business.com, etc.), was given a fake picture, fake name, and a fake expertise profile listing her as a “biochemist and science educator," with 12 years of experience.
Academized’s explanation is that the company guarantees its writers “full anonymity,” and therefore uses ficticious names and pictures. Sure.
Ironically, Rebecca Leigh’s picture also happens to match an expert technology writer listed on the website of a similar media service provider, LeadDev, but this time under the name “Sara Sparrow.” I guess no one wanted to spring for the rights to a new headshot.
What’s being done?
For those not in the media industry, journalists often use companies that provide online networking services to connect journalists with expert sources, especially in a time crunch.
Two of the more popular services, Qwoted, based in New York, and ResponseSource in London, were found promoting many of the fake AI experts profiled by Waugh on their networks.
Since monitoring the sites, Waugh said the companies have reportedly taken action to remove the digital impostors, although he also said ResponseSource did so only after he alerted them.
On a personal note, when recently using Qwoted to request expert commentary on an article about the White House Signalgate scandal and human error in cybersecurity (shameless plug), the service did red flag several “expert” responses to my query – labeling them as possibly AI-generated.
As seen here, the Qwoted warning states: “Wow, that was fast! This pitch came within 15 minutes of your request. Sometimes, a super-fast turnaround like this might mean AI was involved. To double-check, we recommend using Check for AI, powered by GPTZero!”
Quoted provides a “Report” button and the free on-the-spot "Check for AI" feature using GPTZero, which is a standalone AI tool that will check any copy to see if it was written by ChatGPT, Claude, or Gemini chatbots. These buttons are found on every response pitch sent to a journalist.
Quoted’s blog "On Combating Disinformation in the Age of Artificial Intelligence," further states that besides recently adding the advanced AI detection tools, it also uses human moderators to "monitor the authenticity of conversations happening in-platform."
“In a time where accuracy and authenticity are more important than ever, Qwoted is committed to continually investing in tools and techniques that maintain the authenticity of the information on our platform,” the company said.
Waugh says that with “ChatGPT bringing the cost of ‘expert’ commenting down to zero, media services simply need to work harder to stamp out 'fake people.'”
Your email address will not be published. Required fields are markedmarked