Semaglutide, a glucagon-like peptide-1 (GLP-1) receptor agonist, has demonstrated significant weight-loss and cardiovascular benefits in clinical trials, while also sparking widespread public discussion and media attention worldwide (1). Its popularity has been largely driven by social media dissemination, resulting in soaring market demand and global shortages (2). In China, semaglutide quickly became a focus of public debate following its approval for weight-loss indications. On social media platforms, medical professionals display considerable variation in their portrayals of its efficacy and risks, with major perspectives including framing it as a “medical breakthrough,” cautioning against its misuse as a “non-slimming shortcut,” and expressing nuanced attitudes toward patient use (3). These discussions reflect the intertwining of the medicalization of obesity and discourses of personal responsibility in public health communication, while also revealing the communicative dilemma faced by Chinese medical disseminators in balancing promotion and risk warning.

Meanwhile, the enormous popularity of semaglutide has brought challenges concerning information quality and safety risks. In addition to legitimate medications, unregulated synthetic versions have appeared on the market, whose uncertain dosages and potential side effects have raised public concern (2). Against this backdrop, social media has become a primary channel through which the public obtains drug-related information, yet issues of content quality and reliability have become increasingly prominent. At the international level, Campos-Rivera et al. analyzed Spanish-language TikTok videos and found that semaglutide-related content had an average DISCERN score of only 29.8 out of 75, with 71% of videos failing to mention any side effects. Instead, these videos reinforced social notions that attribute obesity to individual responsibility, revealing the potential public health risks posed by the commercial motivations behind the “pharmaceuticalization of obesity” (4). Similarly, Yeung et al. examined 189 English-language YouTube videos and found that both short and long formats generally lacked in-depth discussion of critical issues such as long-term safety, effects of drug half-life, and counterfeit drug risks. Even videos produced by verified medical sources failed to adequately address potential serious adverse events of regulatory and public health interest, such as pancreatitis and gallbladder disease, for which causal associations with semaglutide remain under investigation (5). A cross-platform comparative study by Propfe and Seifert further revealed the significant influence of platform ecology on information quality: YouTube, with a higher proportion of medical professional creators (48%), emerged as the most reliable platform. In contrast, Instagram and TikTok were dominated by non-professional users, resulting in the lowest information quality, with key risk information such as gastrointestinal side effects entirely absent on Instagram (6). The study’s information scoring system showed that average scores across platforms were generally low (YouTube mean 1.38, Instagram mean 0.22), highlighting the systemic shortcomings of social media in disseminating critical drug information.

In the Chinese context, Huang conducted a content analysis of medical professionals on the TikTok platform, revealing a complex narrative tension: semaglutide is simultaneously portrayed as a “medical breakthrough” and cautiously framed as “not a shortcut to weight loss,” demonstrating the interweaving of discourses of medical authority and self-responsibility. This narrative framework not only reflects the communicative dilemma faced by professional disseminators in balancing drug promotion and risk warning, but also exposes the potential influence of “fat phobia” and the moralization of the body in Chinese sociocultural contexts on health communication (3). A synthesis of existing studies shows that common issues in semaglutide-related content on social media include uneven information quality, selective presentation of risk information, and the infiltration of commercial logic into health communication. However, current research mainly focuses on Western mainstream platforms, while systematic evaluations are lacking regarding the patterns of drug information dissemination—especially the reliability and communicative characteristics of side-effect information—on uniquely Chinese short-video ecosystems such as TikTok, Rednote, and Bilibili. Given the significant role these platforms play in Chinese public access to health information, conducting relevant research holds considerable theoretical and practical significance.

Therefore, this study aims to conduct a cross-platform comparative analysis of semaglutide-related short videos on three major Chinese platforms—Bilibili, TikTok, and Rednote. The primary objective is to evaluate and compare the overall quality and reliability of these videos, with a specific focus on how side effects are presented across different platforms and uploader types. The secondary objectives include: (1) characterizing the thematic content distribution of semaglutide-related videos; (2) identifying platform- and creator-specific patterns in the reporting of specific side effect categories (e.g., gastrointestinal, hepatobiliary, ocular); and (3) examining the association between video quality metrics (as measured by modified DISCERN and GQS) and user engagement indicators. By clarifying these dimensions, this study seeks to inform evidence-based strategies for tailoring public health communication to the unique affordances of each platform.

This study provides a comprehensive, platform-specific evaluation of the quality and reliability of semaglutide-related content on three predominant Chinese short video platforms. Our findings reveal significant disparities in content quality, user engagement, and the narrative framing of side effects across Bilibili, TikTok, and Rednote. These results underscore the profound influence of platform ecology and creator identity on the construction and dissemination of public health information.

The hierarchical gradient in content quality—Rednote > TikTok > Bilibili, as evidenced by both M-DISCERN and GQS scores—is a central finding of this work. This gradient can be interpreted through the distinct “affordances” and cultural norms of each platform (25–27). Rednote’s ecosystem, built upon in-depth, community-verified “notes” and a dominant user base actively seeking detailed lifestyle and health information (particularly regarding weight management), creates an environment that rewards evidence-based and well-contextualized content (4, 6). This “pro-knowledge” culture encourages creators to provide higher-quality information to gain credibility and traction. In contrast, TikTok’s core mechanic is rapid, algorithm-driven consumption for mass entertainment (28–30). While this allows health topics to achieve viral reach, the platform’s inherent “attention economy” often prioritizes sensationalism and brevity over depth and nuance, resulting in middling quality scores (31). Bilibili’s position at the bottom of the quality gradient is perhaps the most intriguing. Historically a hub for long-form, niche, and academically-oriented content, our data suggest that its foray into short-form video may have diluted this advantage for complex topics like pharmacology (3, 32, 33). The smaller initial audience for specialized medical content, combined with potential “information dilution” in the short-video format, may lead to a proliferation of low-effort, repetitive, or poorly-sourced videos, as low-quality content can easily fill a knowledge vacuum (34–36).

The critical role of the creator is further elucidated by our data. The significant superiority of professional individuals over non-professional individuals in both quality and engagement metrics highlights that expertise, when effectively communicated, is a key driver of reliable health communication (37, 38). However, the blurred line between professional and non-professional institutions—where significant differences were observed only in engagement metrics, not in core quality scores—warrants caution (39, 40). This suggests that institutional branding alone does not guarantee informational quality. It is plausible that some non-professional institutions (e.g., commercial media outlets) employ production values and narrative techniques that mimic authoritative sources, thereby achieving similar perceived quality scores without the underlying expertise, a phenomenon linked to the concept of “manufactured authenticity” (41, 42).

The high correlation between different engagement metrics (e.g., likes, shares, collects) indicates a cohesive user behavior pattern, where a video’s popularity is reinforced across multiple interactive dimensions (43). The absence of a correlation between publication time and engagement is a noteworthy insight for public health communicators (44). It suggests that for high-value, “evergreen” health topics like drug side effects, content quality and inherent utility are the primary determinants of long-term impact, not ephemeral algorithmic trends at the time of posting. This empowers creators to focus on substantive value rather than tactical timing (45).

Perhaps the most clinically significant finding lies in the “platformization” of side effect narratives. The statistically significant differences in the reporting of specific side effects (e.g., abnormal lab findings, gastrointestinal, and hepatobiliary effects) indicate that a user’s understanding of semaglutide’s risk profile is not based solely on clinical evidence, but is heavily mediated by the platform they use (6, 46). For instance, the emphasis on gastrointestinal side effects on certain platforms may reflect their relatability and potency in personal experience narratives, whereas more abstract “abnormal indicator” side effects might be highlighted on platforms with a more analytical or science-communication orientation (47, 48). This fragmentation of risk discourse can lead to an incomplete or biased public perception, where some risks are over-emphasized while others are minimized, depending on the digital silo one inhabits (4, 49). Our dual-lens analysis provides pivotal evidence for understanding how semaglutide side-effects are communicated on social media. In the full sample (n = 607), conventional statistics masked key differences; for example, no inter-platform disparity was detected for “abnormal laboratory findings”—a toxicity that requires specialist knowledge (p = 0.611). However, when the analytic lens was narrowed to the core sample of videos that actually mentioned adverse events (n = 237), the same comparison became significant (p = 0.036). This phenomenon stems from the fact that 61% of videos in the full cohort contained zero side-effect content, effectively diluting the risk signal. Only after excising this noise could genuine between-platform differences in highly specific, easily overlooked adverse events be detected. Importantly, the “platform effect” for gastrointestinal and hepatobiliary side-effects proved robust: regardless of whether the full or core sample was used, TikTok consistently displayed a higher prevalence of these intuitively perceptible adverse events than Bilibili or RedNote, indicating that the platform’s algorithm or audience preference amplifies the visibility of such salient toxicities. The dual-sample strategy offers four clear advantages. First, the core subset minimizes zero-event dilution, boosting statistical power to uncover subtle platform- or creator-level differences. Second, the approach marries a macroscopic trend (full sample) with microscopic validation (core sample), creating a “breadth-depth” double check that markedly strengthens inference. Third, findings directly inform precision public-health action by pinpointing which high-specialty risk topics (e.g., laboratory abnormalities) are systematically under-represented on which platforms, thereby supplying data-driven targets for platform-specific education or regulation. Finally, the framework is highly transferable: it can be replicated in studies of other drugs or health issues, preserving both population-level generalizability and granular detail while avoiding an “all-or-none” information loss.

These findings have several critical implications. For healthcare consumers and patients, our study serves as a cautionary note, urging critical appraisal of the platform source and creator credentials when seeking health information online. For health professionals and educators, there is a pressing need to actively engage with these platforms, particularly those like Rednote that demonstrate a receptivity to high-quality content, to ensure accurate voices are amplified (50). For content creators, the success of the “strong evidence + scenario-based” expression model on Rednote offers a valuable template for translating complex medical information into accessible yet reliable formats (51). From a policy perspective, our research highlights the need for platform-specific governance. A one-size-fits-all approach to health misinformation is unlikely to be effective. Platforms like TikTok, with their vast reach and entertainment focus, might require more robust fact-checking partnerships, whereas Bilibili could benefit from initiatives that bolster its nascent community of expert creators in the short-video domain (52).

This study is not without limitations. Firstly, its cross-sectional design provides a snapshot in time, and the dynamic nature of social media content necessitates longitudinal follow-up. Secondly, our quantitative analysis, while robust, could be enriched by qualitative methods, such as a discourse analysis of the comment sections, to understand user interpretation and concerns. Thirdly, the sample was limited to three major Chinese short-video platforms, which may constrain the generalizability of our findings to other platforms or cultural contexts. In particular, the study did not include international short-video platforms such as YouTube Shorts, due to substantial ethical, legal, and practical constraints in obtaining and analyzing health-related content across jurisdictions. Consequently, the conclusions should be interpreted primarily within the context of Chinese short-video ecosystems. Despite these limitations, the strengths of this study include a relatively large, systematically sampled dataset, the use of validated instruments (M-DISCERN and GQS) for quality assessment, and a novel comparative framework that explicitly links platform ecology to health information quality and specific risk narratives.

In conclusion, our investigation demonstrates that the landscape of semaglutide information on Chinese short video platforms is not a monolith but a varied ecosystem where quality and narrative are powerfully shaped by the interplay of platform design, creator expertise, and user engagement patterns. Recognizing this “platformized” reality is the first step toward developing more sophisticated, targeted, and effective strategies for digital health literacy and communication. Ensuring the reliable dissemination of pharmaceutical information in the digital age requires not only creating good content but also understanding the distinct soil in which it is meant to grow.

By analyzing 607 semaglutide-related short videos across three major Chinese platforms, this study revealed a consistent quality gradient of “Rednote > TikTok > Bilibili”, underscoring the significant role of platform ecology in shaping information dissemination. Side-effect narratives exhibited a distinct “platformization” pattern, with significant differences in the mention of gastrointestinal and hepatobiliary adverse events across platforms. The dual-perspective analysis further demonstrated that such disparities were more pronounced in the side effect-focused subsample, highlighting how platforms selectively filter risk information. These findings indicate that the drug information accessible to the public is structurally biased. Future interventions must therefore be tailored to platform-specific characteristics to enhance the reliability and completeness of online pharmaceutical content.