Mediahuis suspends senior journalist over AI-fabricated quotes
TL;DR:
- Peter Vandermeersch, the former head of Mediahuis’s Irish operations, has been suspended after publishing dozens of AI-fabricated quotes without verifying them.
- Seven people named in his Substack newsletter confirmed they never made the statements attributed to them, according to an investigation by NRC.
- The case is a pointed reminder that AI hallucinations can bypass even experienced editorial judgement when verification steps are skipped.
Peter Vandermeersch spent decades in senior editorial roles, including a stint as editor-in-chief of NRC in the 2010s and leading Mediahuis’s Irish operations, which publish the Irish Independent. His side project, a Substack blog called Press and Democracy, wrote regularly about journalism’s role in a healthy society.
It was that blog that brought him down. Vandermeersch admitted using ChatGPT, Perplexity and Google’s NotebookLM to summarise reports, then publishing the AI-generated summaries — complete with fabricated quotes — without checking whether anyone had actually said the words attributed to them.
The investigation
NRC, one of Mediahuis’s own titles, found that Vandermeersch had published “dozens” of false quotes. Seven individuals named in his posts told the newspaper they had never made the statements attributed to them. Mediahuis has since removed several of his articles from the Irish Independent website.
In a Substack post headlined “I am admitting my mistake,” Vandermeersch acknowledged the failure directly: “I wrongly put words into people’s mouths, when I should have presented them as paraphrases.”
He added a sentence that read more like a warning label than an excuse: “These language models are so good that they produce irresistible quotes you are tempted to use as an author.”
Why this matters beyond one newsletter
Vandermeersch’s case is uncomfortable precisely because he is not a careless beginner. He is someone who regularly advocated for editorial rigour and described himself as an AI enthusiast who wanted to experiment. Mediahuis CEO Gert Ysebaert said the group applies “strict rules for the use of AI, where diligence, human oversight and transparency are essential.”
The gap between policy and practice is the real story. Newsrooms across Europe are adopting AI summarisation tools at speed, and most rely on individual journalists to verify outputs. When someone with Vandermeersch’s experience fails to check, it raises questions about whether policies alone are enough.
Looking forward
The incident will likely accelerate calls for structural verification requirements rather than guidelines that depend on individual discipline. For any organisation using AI to handle factual content, the lesson is blunt: summaries hallucinate, and confidence in the tool is no substitute for checking the output.