But how good is that advice? A new cross-sectional study published in Frontiers in Digital Health offers the most systematic answer yet, and the verdict is sobering. A team of rehabilitation researchers based in Shanghai and Jinan retrieved 800 candidate videos from the four major Chinese short-video platforms using six keyword combinations pairing dysphagia or swallowing with feeding, diet, or nutrition. After rigorous screening—removing duplicates, irrelevant content, non-Chinese videos, poor-quality recordings, commercial material, and videos with unidentifiable creators—278 videos remained. Each was scored by two qualified rehabilitation therapists using three established instruments: the Global Quality Score (GQS) for overall educational quality, the JAMA Benchmark Criteria for source transparency, and the modified DISCERN (mDISCERN) scale for information reliability.

The results reveal a deep mismatch between what is popular and what is trustworthy. The overall quality of feeding guidance videos was mediocre at best, with a median GQS score of just 2 on a 5-point scale. Professional identity mattered in revealing ways: videos uploaded by healthcare professionals scored significantly higher than those from non-professionals on both source transparency and reliability, with differences reaching statistical significance at P < .001 for both the JAMA and mDISCERN scales. Yet on the GQS measure of perceived overall quality, the gap between professional and amateur content essentially disappeared—a finding suggesting that viewers cannot reliably distinguish sound medical guidance from plausible-looking improvisation.

The study’s granularity exposes how uneven the ecosystem is. Among healthcare uploaders, rehabilitation nurses delivered the strongest overall quality scores, likely because nursing work emphasizes patient education, plain language, and procedural demonstration. Healthcare professionals from related clinical departments achieved the highest reliability ratings, while individual science communicators scored lowest on mDISCERN, reflecting systematic weaknesses in content balance, source citation, and disclosure of uncertainty. Perhaps most striking was the fate of videos posted by people with dysphagia themselves: these clips attracted the highest engagement—a median of 112 likes and 21 comments—but carried the lowest source transparency, with a median JAMA score of just 1 out of 4.

Platform ecosystems proved equally consequential, and the effects were dimension-dependent. Differences in overall quality across platforms were modest; only Bilibili and Kwai differed significantly on GQS. Source transparency showed moderate variation, with Rednote, Douyin, and Bilibili all significantly outperforming Kwai, which imposes weaker requirements on author attribution and credential verification. Reliability showed the starkest divergence of all: Douyin’s videos achieved mDISCERN scores significantly higher than every other platform, with an effect size (ε² = .503) that the authors describe as the most pronounced platform-level difference in the study—likely reflecting Douyin’s creator certification and institutional account mechanisms, which force upfront disclosure of who is speaking.

Beyond quality, the study mapped what the videos actually covered against a ten-domain framework built from clinical guidelines, including recommendations from the American Speech-Language-Hearing Association and the 2023 Chinese guideline for rehabilitation management of swallowing disorders. The coverage map is lopsided in a dangerous way. Food texture adjustment—thickened liquids, appropriate bolus sizes—was the most covered domain at 77 percent, followed by swallowing posture adjustment at 52 percent. These are visually demonstrable, easily replicated techniques, perfectly suited to the short-video format, which explains their popularity.

The safety-critical domains fared far worse. More than 70 percent of videos never mentioned feeding environment adjustment, feeding risk control, oral care, nutritional management, or medication management. Nutritional management appeared in only 21 percent of videos and oral care in 17 percent. Most alarming of all, medication management—the knowledge of which pills can safely be crushed, dissolved, or mixed with food—was covered by a mere 4 percent of the sample. The clinical stakes here are enormous: crushing an extended-release or enteric-coated formulation can trigger dose dumping or drug inactivation, potentially precipitating severe hypotension, respiratory depression, seizures, or death, and medication administration errors are already significantly more common among patients with dysphagia than without.

The researchers also documented a troubling disconnect between virality and veracity. Rehabilitation therapists’ videos were short—a median of 67 seconds—yet earned the most likes and shares, demonstrating that concise, practical professional content can win the attention economy. Conversely, the high-engagement, low-reliability pattern seen in patient-uploaded content echoes earlier findings that interaction traffic and information quality frequently diverge on health topics. The search process itself exposed platform dysfunction: over 58 percent of initially retrieved videos were irrelevant to dysphagia feeding guidance, a failure the authors attribute to algorithmic confusion of medical terms with entertainment content such as eating-challenge videos, keyword semantic drift, and the mismatch between lay search phrases like choking while eating and clinical vocabulary like dysphagia.

The study’s authors are careful about its limits. The sample reflects only what recommendation algorithms surfaced for fresh, history-free accounts in Shanghai over three days in September 2025, and the personalized, dynamic nature of those algorithms makes replication difficult. Cross-platform comparisons of raw engagement counts are complicated by differences in user base size and traffic distribution, and the domain-coverage framework, developed by the team, awaits external validation. The design also captures content, not consequences—causal links between video quality and outcomes such as aspiration pneumonia or unplanned readmissions remain unproven.

Even so, the implications are actionable. The authors call for clinicians to curate and recommend certified video resources as an extension of patient education, for platforms to introduce credential verification, tiered labeling of medically certified accounts, quality-weighted recommendation algorithms, and semantic mapping between everyday and medical search terms, and for the field to adopt multidimensional assessment combining tools like GQS, JAMA, and mDISCERN. The central lesson is one that health communicators have been learning across platforms and diseases: in the short-video era, reach is not reliability, and a confident sixty-second demonstration about something as deceptively simple as a safe swallow can either extend clinical care beyond the hospital wall—or quietly undermine it.

Subject of Research: Quality and reliability of short-video health information on feeding guidance for dysphagia across Chinese social media platforms

Article Title: Quality, source transparency, information reliability, and domain coverage of short video-based information on feeding guidance for people with dysphagia: a cross-sectional study on 4 major Chinese social media platforms

Article References: Chen, J., Xiong, T., Li, Y., Wang, Y., Yang, C., Huang, J., & Fang, F. (2026). Quality, source transparency, information reliability, and domain coverage of short video-based information on feeding guidance for people with dysphagia: a cross-sectional study on 4 major Chinese social media platforms. Frontiers in Digital Health, 8, Article 1919212. https://doi.org/10.3389/fdgth.2026.1919212

Image Credits: AI Generated

DOI: 10.3389/fdgth.2026.1919212

Keywords: dysphagia, feeding guidance, short-video platforms, health information quality, social media, rehabilitation, aspiration pneumonia, medication safety, JAMA benchmark, DISCERN, Douyin, health communication