How good is the information quality of influenza short videos on Chinese platforms?

A 229-video audit finds Bilibili stronger on credibility and TikTok on clarity, with quality largely uncoupled from engagement.

Direct answer

Influenza short videos on TikTok and Bilibili are only moderately reliable, and the two platforms fail in different ways: Bilibili videos scored higher on source credibility and overall quality, while TikTok videos scored higher on understandability and actionability. Across 229 videos, quality scores were largely uncorrelated with likes, comments and shares, so popularity was not a usable proxy for accuracy.

1sources cited

This article was generated with WisPaper-powered search and paper analysis.

Four scales that score presentation, not truth

The study is a cross-sectional audit of 229 influenza-related short videos, collected on October 7, 2025, from TikTok and Bilibili using newly registered accounts and the Chinese keywords for influenza. Only the first 100 results per keyword per platform were collected, because the authors reasoned that users typically browse only the first pages of search results. That makes this a purposive sample of what a typical searcher would encounter, not a random sample of all influenza videos.

Each video was scored on four instruments that measure different things. The JAMA benchmark gives up to 4 points for whether a video names its author, provides attribution, discloses conflicts and shows currency. The Global Quality Score (GQS) is a single 1-to-5 holistic rating of flow, practicality and completeness. Modified DISCERN scores reliability and objectivity, including whether sources are cited and whether information is balanced. PEMAT splits into understandability and actionability, each reported on a 0-to-100 scale, capturing whether language is clear and whether viewers are given specific things to do.

The critical boundary is stated by the authors themselves: none of these tools verifies factual accuracy. A video can be transparent, well organized and easy to act on while still containing a wrong clinical claim. The scores therefore describe communication quality and credibility signals, not whether the medical content is correct.

Doctors and lay creators dominate, treatment leads

Doctors uploaded 73 of the 229 videos (31.9%) and non-professional individuals uploaded 88 (38.4%), making these the two largest source groups. Disease knowledge was the main content in 176 videos (76.9%), and treatment was the most common theme at 85 videos (37.1%), followed by symptoms at 70 (30.6%). Prevention accounted for 44 videos (19.2%), a smaller share than treatment.

The platform mix behind those totals differs sharply. On TikTok, doctors supplied 62 of 135 videos (45.9%) and news agencies 30 (22.2%), while non-professional individuals supplied 25 (18.5%). On Bilibili, non-professional individuals supplied 63 of 94 videos (67.0%) and doctors only 11 (11.7%). Bilibili also carried far more pathogenesis and pathology content (30 videos, 31.9%) than TikTok (2 videos, 1.5%), and far more animation or action formats (46, 48.9%, versus 11, 8.2%).

This distribution matters for reading the quality comparison later: the two platforms are not just different interfaces for the same creators, they host different creator populations and different content mixes.

TikTok wins reach while Bilibili runs longer

TikTok videos were much shorter, with a median duration of 52 seconds against 421 seconds on Bilibili, and were more recent, with a median of 239 days since upload versus 660.5 days. TikTok videos also drew far more interaction: median likes of 1,851 versus 154, comments of 159 versus 8.5, and shares of 1,008 versus 109. Uploaders on TikTok had a median of 347,000 followers, compared with 11,500 on Bilibili. All of these differences were significant at P < 0.001.

Collections were the exception: the median was 500 on TikTok and 231 on Bilibili, but the difference did not reach significance (P = 0.066). That single non-significant metric is a reminder that the engagement gap is broad but not uniform across every interaction type.

The authors interpret the pattern as reflecting platform design, describing TikTok's traffic-driven algorithm as favoring brief, practical and entertaining content while Bilibili places more emphasis on rigorous science communication. That is the authors' explanation of the pattern, not a tested causal mechanism in this dataset.

Credibility favors Bilibili, usability favors TikTok

On the credibility-oriented scales, Bilibili scored higher. The median JAMA score was 2.5 (interquartile range 2 to 3) on Bilibili versus 1 (1 to 2) on TikTok, and the median GQS was 4 (4 to 4) versus 3 (3 to 4); both differences had P < 0.001. Modified DISCERN medians were 3 in both groups, with a smaller but still significant difference (P = 0.012).

On the communication-oriented scales, the direction reverses. TikTok's median PEMAT understandability was 100 (82 to 100) against 71.5 (64 to 90) on Bilibili, and median actionability was 67 (0 to 67) against 33 (0 to 67), both P < 0.001.

These are not contradictory findings, because the scales answer different questions. A short TikTok video can be plainly worded and end with a concrete instruction while omitting authorship, citations and balanced caveats; a longer Bilibili explainer can cite sources and cover uncertainty while burying the practical takeaway. Reading one score as a verdict on the other would misstate what the study found.

Source type and topic track with higher scores

Videos from physicians, news agencies and non-profit organizations, and videos focused on treatment or diagnosis, generally showed higher quality than those from non-professional individuals and for-profit organizations. The authors read this as an argument for stronger professional participation in health content creation and oversight.

The paper also reports a recurring confusion encountered during screening: many videos labeled as influenza were actually about COVID-19 or the common cold and were excluded under the predefined criteria. The authors treat this as evidence that creators and platforms conflate influenza with other respiratory infections. Among the small share of videos focused on diagnosis (8 videos, 3.5%), some medical education content specifically explained how to distinguish influenza from COVID-19 or the common cold and stressed viral testing rather than self-diagnosis.

That exclusion is worth noting when interpreting the sample: the 229 analyzed videos are those that survived a topic filter, so the measured quality applies to videos that were genuinely about influenza, not to the broader pool of mislabeled respiratory content.

Engagement barely tracks quality, and differs by platform

The four quality scales were positively correlated with one another on both platforms (all P < 0.05), which supports treating them as related but distinct dimensions. The strongest TikTok correlations were between JAMA and modified DISCERN (r = 0.57) and between GQS and modified DISCERN (r = 0.54); on Bilibili the strongest was GQS with modified DISCERN (r = 0.59).

Correlations between quality scores and engagement were weak and inconsistent. On TikTok, most interaction metrics showed no significant correlation with ratings; the notable exception ran the wrong way, with comments mildly negatively correlated with GQS (r = −0.30) and modified DISCERN (r = −0.22). Follower counts showed no significant correlation with any rating on TikTok.

On Bilibili the pattern leaned positive: modified DISCERN correlated with likes (r = 0.36), collections (r = 0.39) and shares (r = 0.36), and uploader follower counts correlated with all ratings except actionability (r = 0.26 to 0.44). The authors conclude that on TikTok high dissemination does not indicate high-quality information, while Bilibili shows a relatively more rational information ecosystem. These are correlational patterns from a single snapshot, not evidence that engagement causes or rewards quality.

A snapshot of two platforms, not a verdict on accuracy

Several design features bound what these results can support. Only TikTok and Bilibili were sampled, using top search results rather than a random draw, so the findings do not generalize to other platforms or to all influenza videos. The data are a single cross-sectional snapshot as of October 7, 2025, and video content and rankings change over time. Ratings used objective criteria but retain subjective elements, and the instruments cannot verify factual accuracy, so the study did not fact-check claims against clinical guidelines and could not quantify how much inaccurate content circulates.

Within those bounds, the paper's practical argument is platform-specific rather than uniform. The authors suggest that TikTok may need concise, behaviorally focused videos while Bilibili may need detailed, evidence-oriented explanatory content, and that platforms should build basic quality criteria into recommendation algorithms and clearly label verified medical sources.

The most decision-relevant takeaway is the mismatch the data expose: the platform with the strongest reach and engagement, TikTok, scored lowest on source credibility and overall quality, and its engagement metrics carried little or negative association with quality. For anyone using these platforms to judge influenza information, the paper's evidence supports treating view counts and likes as signals of reach rather than of reliability, and checking authorship and sourcing instead.

Sources used in this answer

1

Evaluating influenza-related short videos on Chinese platforms: information quality, user engagement, and public health implications

Across 229 influenza short videos, Bilibili scored higher on JAMA and GQS while TikTok scored higher on PEMAT understandability and actionability, and quality scores were weakly or negatively correlated with engagement metrics.