Nobody has measured it properly yet. The most specific figure available comes from a 2024 BBC World Service investigation, where a panel of health experts reviewed 23 health-focused episodes of a podcast the BBC described as the UK's number-one ranked show, found potentially harmful claims in 15 of them, and counted an average of 14 claims per episode across those 15 that contradicted established scientific evidence. That investigation was not peer-reviewed. Academic scoping reviews say something narrower and more uncomfortable: the accuracy of user-generated health podcasts is unknown, and studies rarely document how their podcasts were chosen or quality-checked.
In 2024, BBC World Service journalists reviewed 23 health-focused episodes of a podcast the BBC described as the UK's number-one ranked show, with a panel of health experts assessing the claims. Fifteen of the 23 contained potentially harmful claims, and across those 15 the average was 14 health claims per episode that contradicted established scientific evidence. Claims that a ketogenic diet can treat cancer. That autism can be "reversed" through diet. That prescribed medications are inherently toxic. All broadcast without meaningful challenge.
That investigation was not peer-reviewed, and its methodology, structured as it was, cannot be treated as the equivalent of a systematic academic content analysis. It was specific and it was expert-reviewed though, and it points at a problem the academic literature has been quietly documenting from a different direction. Two scoping reviews of health podcasts come at this from different sides. One, covering 11 studies, reports podcasts as effective for behaviour change. Both show how little is known about what goes into them: the accuracy of user-generated health podcasts is unknown, and studies rarely explain how their podcasts were chosen or quality-checked. So the medium is highly persuasive, and it gets remarkably little scrutiny for something reaching an audience that size.
The things that make a podcast good at changing what someone does are the same things that make it good at spreading something false. You do not get one without the other.
What the research actually shows
Large-scale quantitative audits of health claim accuracy across major podcast catalogues essentially do not exist in the academic literature. What does exist is a body of evidence approaching the problem from next door: scoping reviews on health-education podcasts, systematic reviews on health misinformation across digital platforms, and a Brookings Institution machine-learning analysis of political podcasts which found more than one in ten episodes carrying potentially false information. Those episodes collectively picked up over 100 million engagements.
The systematic review evidence on health misinformation in social media (Suarez-Lledo and Alvarez-Galvez, 2021) gives the most granular category breakdown available, though it covers social media broadly rather than podcasts specifically. Misinformation runs highest in smoking and drug content, where some studies found up to 87% of sampled posts carrying it. Vaccine content came next, at around 43% of sampled posts in some studies. Smoking, drugs and vaccines are among the topics that recur across popular health and wellness podcasts.
The misinformation prevalence figures above come from studies of social media content, not from podcast-specific audits. Applying them straight to podcasts would be an overreach. They are the closest available benchmark and they are cited as exactly that. The gap in the literature itself, meaning the absence of any large-scale quantitative audit of podcast health claims, is a meaningful data point on its own.
Why podcasts are structurally prone to this
There is nothing mysterious about the mechanism. Podcasts are long, conversational, and mostly listened to while doing something else, so during a workout or a commute or while cooking. Few people in that position stop to check a claim as it goes past. The format rewards keeping the story moving over getting it exactly right, and when a host or a guest says something compelling and slightly wrong, cutting it leaves a hole in the audio, and the hole usually wins.
The money makes it worse. Independent podcasts get monetised through advertising, sponsorships and affiliate links, and the content that drives downloads and gets clipped and shared tends to be surprising or counterintuitive or extreme. "The established view is basically right" does not produce clips. "What you've been told about [X] is wrong" does. That is not individual hosts failing morally. It is a reasonable response to how attention-based media pays.
And the format is an intimate one. The Brookings analysis, citing work on podcasting as an intimate medium, argues that the closeness between a podcaster and their audience may make listeners more likely to believe untruths.
The reach numbers matter here
Edison Research's 2024 Infinite Dial survey, cited in a JMIR scoping review, reported that 47% of Americans aged 12 and over listen to a podcast at least monthly, against more than five million podcasts registered globally as of 2023, most of them no longer publishing. That same scoping review pulled together 51 studies and reports on podcasts used in health interventions. Most sat inside multicomponent programmes aimed at physical health, and only seven of the 26 published studies tested a podcast on its own. So the evidence that podcasts change health behaviour comes from a small and young body of work. If they do change what people do, that cuts both ways.
Whatever makes an evidence-based health podcast genuinely useful is the same thing that makes a misleading one genuinely harmful. The format is powerful either way. What you put into it decides whether that power helps the people listening or damages them.
What responsible hosting looks like in practice
The academic and professional literature converges on a few practical principles, consistent enough across sources to be worth stating directly.
Treat uncertainty as information. Claims want qualifying to match what the evidence actually supports, so a mechanism that is plausible but unproven gets flagged as exactly that rather than delivered as established fact. Sounding certain where the evidence is not does not persuade an informed listener. It costs you with them.
Guest claims want scrutiny before the broadcast rather than after it. Asking for references in advance, and being willing to push back on an unsupported statement while it is being said, is a production decision, and treating it as an editorial nicety is how it gets skipped. A host who lets a guest assert that some supplement reverses a chronic condition without pushback has not stayed neutral. They have endorsed the claim in front of every listener who cannot tell the difference between "the guest said this" and "this is true."
Show notes work best as a reference document. A curated list of links to the peer-reviewed sources under the episode's main claims takes real work, and that work is the thing that lets a listener tell your show apart from one that holds itself to nothing. It also happens to be searchable text sitting on your episode page, which helps people find you.
Corrections should be explicit, not implicit. When something from a previous episode proves wrong or significantly overstated, a dedicated correction in a subsequent episode is more credible than simply never mentioning it again.
The platforms have started responding to the most visible cases, and Spotify did add advisory labels for COVID-19 content in January 2022 after public pressure. Podcast-specific enforcement stays inconsistent though, partly because episodes are usually hosted somewhere else and syndicated in rather than published natively. So the editorial burden sits with the host, and there is nothing on the horizon suggesting that changes.