Press Gazette’s tracker of AI journalism scandals records a cluster of cases in which AI-generated writers and sources fooled outlets from Science magazine to Canadian newspapers, exposing gaps in editorial checks
A Cluster Of Deceptions
The past few months have produced an unusual concentration of cases in which artificial intelligence was used to deceive news organizations. Press Gazette maintains a live tracker of AI journalism scandals and mistakes, and its recent entries read like a catalogue of verification failures. Fake writers with AI-generated headshots, fabricated experts quoted dozens of times and a state-linked influence operation placing articles under false names have all reached publication.
The Iranian Bylines
The most serious case, reported by The Washington Post on October 9, involved ChatGPT accounts of Iranian origin that placed almost 100 AI-assisted articles under seven fake bylines in small and mid-sized outlets, mostly on international affairs. One Florida weekly ran a column by a fake author with an apparently AI-generated photograph, after the piece arrived through an existing relationship with a think tank. Middle East Monitor, Daily Kos and the Port St Joe Star took stories down. According to the tracker, the same column remained live at another outlet.
The case shows how generative AI can lower the cost of influence operations. Producing plausible opinion articles in fluent English once required skilled writers. Now a small team can generate dozens of pieces and distribute them through intermediaries that lend them credibility.
The Fake Freelancer
In September, Science magazine discovered that an impersonator posing as a real UK freelance journalist, Caroline Williams, had pitched and filed an article that appears to have been written with AI. Editors communicated with the impersonator only by email, and quotes in the article were later found in other writers’ published stories. Science removed the article. Its managing news editor said simply: we were conned. The magazine is trying to recover the 1,000 pound fee.
The Daniel Robson Persona
In Canada, an apparently AI-generated writer called Daniel Robson, with an apparently AI-generated headshot, placed more than two dozen pieces in Canadian outlets. He avoided in-person verification and benefited from what the tracker describes as a trust cascade: earlier bylines in other outlets made him seem credible to the next editor. The Montreal Gazette, the Ottawa Citizen, the Western Standard, The Hub, Open Canada and Policy Options removed stories. Policy Options said it regretted not verifying him more thoroughly. The Ottawa Citizen now requires in-person meetings and photographs for new opinion writers.
The Expert Who Was Not There
Another case involved a supposed art therapist, Dr Eleni Nicolaou, who was quoted on psychology topics more than 30 times in recent months by outlets including Tom’s Guide, Forbes, Vice, Glamour and Yahoo. Her practice could not be found online, and she appears to be fabricated. The journalist response service Qwoted removed her profile. The case exposes the risks of expert-matching services, which connect journalists with sources at speed, often without independent verification of credentials.
Problems Inside Newsrooms
Not all the recent cases involve outside deception. At Cleveland.com, an AI-assisted story was published under a reporter’s byline even though she had not written or reviewed it. The byline was changed to a generic Express Desk credit with a correction note. Elsewhere, Futurism reported that Brown Brothers Media used fake bylines and false credentials to publish large volumes of AI-generated articles on sites it had bought after laying off writers. The company said it stopped using those bylines and changed its AI policy.
Opinion sections have also faced questions about undisclosed AI use by prominent contributors. In August, the Financial Times added an editor’s note after a Harvard professor used AI to condense a column in breach of its code of conduct, while the Wall Street Journal’s opinion editor defended an op-ed by a hedge fund billionaire that the author confirmed was written with AI help.
Why Verification Is Failing
Several factors make these deceptions easier. Newsrooms have fewer staff, so editors have less time to verify contributors. Many opinion and freelance relationships are conducted entirely by email. Expert databases and pitching services prioritize speed. And AI tools now produce text, images and personas that are convincing at a glance. The trust cascade identified in the Canadian case is especially dangerous: each publication that accepts a fake contributor makes the next acceptance more likely.
What Newsrooms Can Do
The responses already adopted by some outlets offer a guide. Require video or in-person meetings with new contributors. Verify identity documents for paid freelancers. Check that experts’ credentials and practices exist independently. Search quotes to confirm they are original. Be wary of unsolicited pitches arriving through intermediaries such as think tanks or PR agencies. And establish clear policies on AI use by staff and contributors, with disclosure where AI has been used.
There is also a role for the platforms and services that sit between journalists and contributors. Expert-matching services, freelance marketplaces and syndication networks can introduce identity checks and credential verification at their end, rather than leaving every newsroom to repeat the work. Industry bodies could share information about known fake personas, much as cybersecurity firms share threat intelligence.
Trust Is The Product
Journalism’s central promise is that what it publishes is true and that the people named in it are real. Each of these cases breaks that promise in a small way, and together they suggest a systemic vulnerability. As AI tools become more capable, the cost of deception will keep falling. Verification, a basic craft skill that has sometimes been squeezed by cost-cutting, is becoming one of the most important investments a newsroom can make.