In three of my own episode briefs from August, the check caught four things. There was a real meta-analysis credited to an author who isn't on the paper, and a TikTok count whose only source was a wellness summit's own trend post. There was also a burnout number credited to Deloitte that's actually Aflac's, from a 2024 report Aflac has updated since. And there was a trial where two of the authors wrote a book on the exact method they tested, while the paper says they have no competing interests.
One of my research briefs in August cited a big meta-analysis in Nature Human Behaviour, 183 randomised trials, from January 2026. The paper exists, its numbers were the ones in the brief and the link went straight to it, but the brief called it Cain et al. and there is nobody called Cain on that paper. It's Wilkie and colleagues.
Read in a script, nothing in that line would make you stop, because everything you'd normally look at is in order. You find it when you open the paper and look at who actually wrote it.
So what does podcast fact checking actually catch?
You might picture it as catching the claim that sounds too good, the miracle number or the study nobody can find. That can happen too. But the ones from August were much more subtle than that, and that’s where my custom-built mechanisms come in.
They come from three briefs for episodes of my own show that aren't out yet, so everything here is from the briefs, and what the episodes argue can wait until they're out. All four got caught before anything was recorded.
The right paper under the wrong name
That's the Cain one. It's tricky because it comes with more detail than a vague claim would, the journal, the date, the number of trials, and with all that in front of you, you'd assume somebody already checked it. On my show a study is often described on air instead of named, and a credit like the Cain one is part of why.
And a number only one source had
My brief on ambition and burnout used roughly 230,000 TikTok videos under #nervoussystemhealing to show how big that trend had got. The check could trace the figure to one place only, the Global Wellness Summit's own trend post from April 2026, so the people writing up the trend were also the only source for that number. Fast Company had counted 178,500 videos under the same tag in February 2026. Both can be right, of course, since the two were published about two and a half months apart and a count like that can grow in that time, but neither of them tells you how many people actually do anything for their nervous system, they count videos. The script for that episode doesn't give a count for that hashtag now.
A Deloitte number that's actually Aflac's
My brief on productivity and self-worth had 66% of millennials against 39% of baby boomers as a Deloitte finding, taken from coverage of Deloitte's survey, and to be fair the brief had already flagged that line as uncertain. It's Aflac's number, from their WorkForces Report of November 2024, a survey of U.S. workers. And what it counts is moderate to high burnout, which is wider than being burned out. By Aflac's next report, in October 2025, Gen Z had moved ahead of millennials.
A trial where two of the authors wrote the book on what it tested
My brief on awe leaned on a 2025 randomised trial its authors call the first trial of its kind, and the numbers in the brief matched the paper. The practice it tested is spelled out in the paper as Attention, Wait, Exhale and Expand. Two of the co-authors wrote a published book on exactly that method, and the paper's declarations say the authors have no competing interests. The brief didn't mention the book. I wouldn't throw the study out over that. In the script for that episode, the book comes up the first time the trial does, together with the paper saying there are no competing interests.
And why does it have to happen before the recording?
If a claim gets caught in the brief, I change a few words and that's it. After recording it's an edit or a re-record, and if nobody notices at all, it just stays in the published episode. If you have a licence attached to your name, that last case is the one you'd want to avoid.
That's what I mean on my site by the Credibility Shield, every factual claim in the research brief traced to a verifiable source before the host records. The research briefs it runs on are something I've written about before.
I'm not assessing anyone's medicine. I look at the sourcing, and the clinical call stays with the host.
I also have no data on whether shows that check their claims before recording get corrected less often, sued less often or grow faster, and I don't know of anyone who has compared that. Four is a small sample, and what checking does for a show's numbers I can't tell you, but every brief for my own show goes through it now, and it's part of every package I offer clients. In that case, “better be safe than sorry” is the right call.