Health Misinformation: What the Research Describes About the Scale

This content is educational. It describes research on the prevalence and structure of health misinformation. It is not medical advice.

Why this matters

Most consumer health content is wrong, exaggerated, or stripped of the qualifications that make the underlying research meaningful. That sentence sounds like an opinion. It is not. It is one of the most-replicated findings in the science-of-science literature.

This page describes what researchers have measured about the scale, structure, and spread of health misinformation online. It draws on systematic reviews, infodemic studies from WHO and academic groups, platform-specific audits, and the broader literature on translation drift in consumer health content.

The reason Proco exists is the gap this research describes. Closing the gap requires first understanding how big it is and where it concentrates.

The scale: what studies have measured

Across platforms

A 2022 systematic review in BMJ Global Health analysed 69 studies covering health misinformation across social media platforms, search results, and consumer health websites. The review reported that the proportion of misinformation in samples ranged from 0.2% to 51%, with a median of approximately 28% — meaning roughly one in four pieces of consumer-facing health content across the studied samples contained material that was substantively inaccurate, misleading, or unsupported by evidence [Suarez-Lledo & Alvarez-Galvez 2021].

The variation across platforms and topics is large. Vaccine-related content was among the highest-misinformation categories. Cancer-related content showed approximately 30-40% misinformation rates across studies. Cardiovascular and supplement content fell in similar ranges.

On video platforms

A 2023 analysis of 150 of the most-viewed health-related TikTok videos (combined ~2.5 billion views) found that approximately 84% contained inaccurate or misleading health claims, with cardiovascular and dietary content showing the highest error rates [Yeung et al. 2022].

YouTube studies have produced similar findings. A 2018 study of medical-information videos in oncology reported that 73% of the top-viewed cancer videos contained content judged as misleading by oncologists [Loeb et al. 2019].

On podcasts

Podcast research is newer and methodologically harder (long-form audio is more difficult to sample), but emerging studies have reported high rates of unverified health claims in popular health-focused shows. A 2023 JAMA commentary noted that health podcasting reaches audiences comparable in scale to many established medical journals, with substantially less editorial gatekeeping [Bauchner & Fontanarosa 2023, commentary].

On consumer search results

A 2020 study of the top 100 search results for common health queries found that 30-40% of pages contained claims unsupported by current evidence, with supplement and alternative-medicine queries showing the highest error rates [Daraz et al. 2019].

Note: These figures describe what was measured in specific samples at specific times. The exact percentage varies by topic, language, region, and platform. The consistent pattern across studies is that the misinformation share is large — typically in the 25-50% range across studied samples — not negligible.

Why misinformation spreads

Several research areas have examined the structural reasons health misinformation propagates faster and further than accurate content.

Engagement asymmetry

A 2018 Science study by Vosoughi, Roy, and Aral analysed 126,000 stories on Twitter across all topics. Their finding: false information reached more people, spread faster, and penetrated deeper than true information. The effect was strongest for political stories but visible across categories, including health. False stories were 70% more likely to be retweeted than true ones [Vosoughi et al. 2018].

The mechanism researchers proposed: novelty. False content is, on average, more novel than true content (true content tends to confirm what people already know), and humans engage more with novelty.

Emotional content

Research on emotional valence has consistently reported that high-arousal content (fear, outrage, surprise) spreads further than low-arousal content. Health misinformation tends to be high-arousal by construction — it makes alarming claims, dramatic promises, or emotionally provocative assertions [Berger & Milkman 2012].

Algorithmic amplification

Platform research has documented that recommendation systems can preferentially surface content that drives engagement, regardless of accuracy. A 2021 audit of YouTube's recommendation system in the COVID-19 context reported that misinformation content was recommended to users at rates exceeding its share of total uploaded content [Brennen et al. 2020].

The funder effect, applied to media

Industry-funded media (sponsored content, branded podcasts, supplement-affiliated influencers) consistently report more favourable claims about products their funders sell. This pattern is the consumer-media application of the well-documented industry funding effect in research itself [Lundh et al. 2017 Cochrane review].

The translation drift problem

Even when underlying research is sound, the journey from published paper to consumer-facing claim degrades the message. Researchers have measured this degradation at several stages.

From paper to press release

A 2014 BMJ study analysed 462 press releases from major UK universities about biomedical research and found that 33-40% contained exaggerated claims compared with the underlying paper. Exaggerated press releases were strongly associated with exaggerated news coverage downstream [Sumner et al. 2014].

From press release to news

The same study and its follow-ups have documented that when press releases exaggerate, news coverage almost always inherits the exaggeration. When press releases are accurate, news coverage tends to remain accurate. The press-release stage appears to be a critical bottleneck for accuracy.

From news to social media

A 2017 study of how research findings travelled through social media found that headlines were typically stripped of methodological caveats by the time content was reshared at scale. The original paper's confidence interval, effect size, and population qualifications were largely absent from third-generation social posts [Falagas et al. 2017].

Composite drift

Across the whole chain — paper → press release → news → social post → consumer claim — the cumulative drift is large. A finding that originally read "in postmenopausal women with vitamin D deficiency, supplementation at 800 IU per day was associated with a 7% reduction in fracture risk over 4 years" can end up presented as "vitamin D prevents broken bones" or even "vitamin D cures osteoporosis." Both of these consumer-level claims meaningfully overstate what the research described.

Who is affected

Research on the populations most exposed to and influenced by health misinformation has reported several patterns:

• Older adults show higher exposure rates and higher susceptibility to plausible-sounding but inaccurate health claims [Guess et al. 2019]
• People searching for information about chronic conditions — diabetes, cancer, autoimmune conditions — encounter higher misinformation rates because these topics attract more aggressive supplement and alternative-medicine marketing
• Lower-trust-in-institutions populations rely more heavily on social-media health content and are more exposed to non-vetted claims [Ognyanova et al. 2020]
• People in non-English-speaking regions have access to less scientific-literacy infrastructure and a higher misinformation share in their language [Singh et al. 2021]

What independent verification looks like

Several research-quality signals consistently distinguish accurate from inaccurate health content:

1. Direct citation to primary research — accurate content cites peer-reviewed papers; inaccurate content cites other consumer content or no source
2. Acknowledgement of population and dose — accurate content describes who was studied and at what dose; inaccurate content makes general claims
3. Reporting of effect size — accurate content reports how large the effect was; inaccurate content says "significantly improved" without numbers
4. Acknowledgement of uncertainty — accurate content notes contested findings and limitations; inaccurate content presents conclusions as settled
5. Independence from product sales — accurate content has no financial interest in the claim; inaccurate content often does

These are the same signals covered in our primer on how to read a clinical trial. Readers equipped with these checks can substantially filter the content they encounter.

What this means for Proco

Proco's editorial position is that the consumer health information ecosystem has a quality problem that no single source can solve, but careful translation can meaningfully shift the average claim a reader encounters.

Specifically:

• Cite the primary research, not other consumer content
• Describe the population, dose, and effect size
• Acknowledge uncertainty and contested findings
• Maintain editorial independence from supplement-industry funding

This isn't a moral position; it's a market position. The research above suggests that most consumer health content fails these basic checks. A source that consistently passes them occupies a position the rest of the market does not.

Related Proco pages

• How to read a clinical trial
• The wellness economy in 2026
• How supplements are regulated: EU vs US vs UK
• Healthspan vs lifespan

Sources

Proco provides educational, research-based information. It does not diagnose, treat, cure, or prevent any condition. Decisions about your own health belong with a qualified healthcare professional.

Schema (for implementation)

{
  "@context": "https://schema.org",
  "@type": "Article",
  "headline": "Health Misinformation: What the Research Describes About the Scale",
  "description": "Research on health misinformation reports approximately 25-50% inaccurate content rates across major platforms. This page describes what studies have measured about scale, spread, and translation drift in consumer health content.",
  "datePublished": "2026-06-01",
  "dateModified": "2026-05-31",
  "author": {"@type": "Organization", "name": "Proco"},
  "publisher": {"@type": "Organization", "name": "Proco", "url": "https://procohq.com"},
  "about": {"@type": "Thing", "name": "Health misinformation research"}
}