Where health myths live: what the evidence says platform by platform

Asking "which platform is worst for health misinformation" sounds like a question with a clean answer. The published evidence is messier than that, and the messiness is itself useful.

A 2026 review by Carter and colleagues in the Journal of Social Media Research examined 27 studies covering 5,057 posts on mental health and neurodivergence content. Measured misinformation rates ranged from 0% to 56.9% depending on the study, topic and method, and were higher on TikTok than on YouTube (https://jsomer.org/index.php/pub/article/view/84). That range is not a failure of the research. It reflects how much the answer depends on what you sample and how you define the error.

A 2024 JMIR scoping review of 70 studies reached a similar conclusion at a broader level: prevalence is significant but highly variable, and the field needs standardised measurement before cross-platform comparisons can be trusted (https://www.jmir.org/2024/1/e38786/). Anyone quoting a single headline percentage for "social media" as a whole is quoting past what the evidence supports.

Where the evidence is firmer is on patient-level exposure. A 2023 JMIR Cancer study found 56% of cancer patients had encountered misinformation about their condition, and 71% said they believed it at least sometimes (https://cancer.jmir.org/2023/1/e43749). Exposure is not an abstract problem for a vulnerable audience making live treatment decisions.

There is also evidence that misinformation can act as an early indicator rather than only a consequence. A 2025 cross-platform study of antimicrobial resistance content found 19.3% of posts contained misinformation, spreading around 2.4 times faster than accurate content, and preceding resistance trends by six to nine months (https://pubmed.ncbi.nlm.nih.gov/41789509/). If that pattern holds, monitoring narratives is not only defensive work — it is a leading signal.

Our Channel Analysis page holds the per-platform detail behind these findings, with each figure linked to the study it comes from. Where a platform has no reliable published measurement for a topic, we say so rather than filling the gap with an estimate.