Two Chinese platforms, Chinese-language videos, one day in February. That population is not the feed a US reader opens and the advertising rules around it are different, so the numbers below travel less well than the pattern they describe — which is worth reading for the same reason what a seller leaves off a page is worth counting.
Three instruments, two answers
Two reviewers scored every video on three established tools. [1] The Global Quality Scale asks how useful a piece of health content is. The modified DISCERN asks how reliable it is. The JAMA benchmark asks four narrower questions: who wrote it, what sources it drew on, what the author’s interests are, and when it was last updated.
Median GQS was 4.0 and median mDISCERN 4.0, both near the top of their ranges. Median JAMA score was 2.0 out of 4. The two platforms were indistinguishable on usefulness and reliability and not on transparency, where the TikTok videos scored a median 3 against Bilibili’s 1, p < 0.001 — so whatever drives the transparency gap is a property of the platform rather than of the subject.
Who made it predicted how good it was
In the multivariable analysis, videos uploaded by medical professionals scored higher on all three instruments. Commercial uploads scored lower on both GQS and mDISCERN.
That is the single most consistent finding in the paper, and it is the one that survives translation to any platform: the uploader’s identity predicts the quality, which is why the benchmark that asks for the uploader’s identity is the one worth caring about.
Engagement ran the wrong way, unevenly
More likes was independently associated with lower reliability and lower quality scores overall. On the TikTok side, more comments tracked lower quality too.
It was not uniform: on Bilibili, more shares went with higher scores. So “popular means worse” is too strong — what the data support is that engagement metrics do not track quality, which the authors say in exactly those terms. A number that rises for unrelated reasons is not a ranking, the same problem as any ordered table whose order is not the thing you want.
What is missing from the feed
82.2% of the videos were about weight loss and 2.7% were about glycemic control — a drug developed for type 2 diabetes, discussed almost entirely as a weight drug.
That is a fair reflection of why most people search for it, and it also means the indication, the trial populations and the reasons somebody might not be a candidate are largely absent from the medium most people meet it in. Those are the same items that go missing from a product page — the ones this site keeps a running count of, and the reason who publishes what is measurable at all.