All data were sourced from publicly accessible TikTok social media content. Research activities complied strictly with the platform’s terms of service and established ethical guidelines for social media research. No personally identifiable information was collected, and all analyses were conducted on aggregated data to preserve user anonymity.

The data used in this study were obtained from the “Aisou Jisou” Index Analysis System [27], which provides TikTok-related data (including both TikTok and the “lite version” of TikTok). The system holds a Chinese software copyright registration number from the National Copyright Administration of China, with the registration number being 14092309. The keyword “Mpox” was used for data extraction. The starting point was set to a month before the first reported mpox case in the nonendemic United Kingdom in May 2022, and the endpoint was set to 2 months after the confirmation of the first clade Ib mpox case by the Chinese Center for Disease Control and Prevention in January 2025. The total duration was 3 years, covering the period from April 1, 2022, to March 31, 2025.

Data retrieval involved the following steps: (1) examination of open application programming interfaces, focusing on core endpoints such as “Search Index,” “Associated Search Terms,” Target Group Index (TGI), etc; (2) development of an automated data acquisition program using Java and Python integrated with HTTP/HTTPS protocol, namely “Aisou Jisou” Index Analysis System. This system used browser automation tools (eg, Selenium) and headless browser technology to simulate user interactions for automated monthly API calls; (3) application of reverse engineering and packet analysis techniques for decryption and parsing of potentially encrypted data; and (4) implementation of rigorous data cleaning and preprocessing, including deduplication, missing value handling, data normalization, and outlier detection to ensure data integrity for subsequent analysis.

The retrieved dataset comprised: Search Index, Associated Search Terms, and the TGI stratified by gender, age group, and provinces in mainland China. The Search Index is a proprietary metric developed by TikTok to assess the popularity of specific keywords. It is derived by weighting the search volume of the keyword and its related content. It should be noted that the Search Index does not reflect the exact search volume. The TGI reflects the relative strength or weakness of the consumption behavior of a specific user demographic for a given keyword, as compared to the overall platform user base [28]. A TGI >100 signifies that the proportion of users within the specified demographic engaging with the keyword exceeds the platform average.

Before LDA topic modeling, data underwent preprocessing to ensure quality: removal of stopwords (eg, modal particles or meaningless symbols), standardization of synonyms (eg, uniform expressions for countries or disease abbreviations), phrase merging (eg, combining multiword terms into single tokens), and word segmentation. These procedures minimized noise and enhanced semantic clarity for topic modeling.

Descriptive analysis of the Search Index was conducted using WPS (Kingsoft Corporation). LDA topic modeling and associated workflows—including data preprocessing, word segmentation, perplexity and coherence evaluation, topic modeling, and visualization—were conducted using Python (version 3.12.0; Python Software Foundation) with the following libraries: “matplotlib,” “numpy,” “jieba,” “scipy,” “gensim,” “pandas,” “pyLDAvis,” “openpyxl,” and “pyecharts.” The “jieba” library is a robust Python tool for Chinese text processing, offering efficient word segmentation, multiple segmentation modes, custom dictionary support, part-of-speech tagging, and keyword extraction capabilities.

Determining the optimal number of topics is critical in LDA topic modeling, as both insufficient and excessive topics can compromise interpretability. Model performance was assessed using perplexity and coherence scores. Lower perplexity values indicate superior predictive performance on unseen data, whereas higher coherence scores correspond to more semantically coherent and interpretable topics [29,30].

For the LDA topic modeling, a corpus of 25,672 text data underwent preprocessing (stopword removal and part-of-speech-based keyword filtering). Subsequently, the LDA topic modeling was applied for topic mining with the following parameters: alpha=0.0125, beta=0.01, iterations=100, chunksize=400, and update_every=1. Model evaluation encompassed perplexity, coherence, and intertopic distance to ensure high topic coherence and interpretability.

Models with topic numbers ranging from 1 to 15 were evaluated. For each model, the top 15 words per topic were identified based on the conditional probability P(w|z) (probability of word “w” given topic “z”). Topic term distribution and overlap were visualized using “pyLDAvis.” A nonoverlapping term distribution across topics served as an additional selection criterion. Separate LDA topic modeling was conducted for peak months of public attention (identified via Search Index peaks) to track topic evolution and thematic shifts over time (see Figure 1). Figure 1 illustrates the study methodology.

From April 1, 2022, to March 31, 2025, the Search Index for “Mpox” on TikTok in China showed 4 significant peaks, corresponding to May 2022, July 2023, August 2024, and February 2025 (see Figure 2). These spikes aligned with major epidemic events both globally and domestically.

TGI analysis revealed fluctuating public engagement across genders without a clear long-term bias; however, male users maintained higher TGI values for extended periods as the outbreak progressed. In terms of age, younger users consistently demonstrated higher TGI, with the 18‐23 years age group exhibiting the highest and most volatile engagement, peaking in late 2024; the 24‐30 years age group followed with moderate but stable attention; while older groups (31‐40 years, 41‐50 years, and 51+ years age group) showed lower and more stable engagement. Spatially, provinces such as Beijing, Tianjin, and Jilin recorded the highest cumulative TGI values of 5922.38, 5692.41, and 3579.90, respectively; individual months also showed sharp increases in regions like Gansu (December 2023) and Tianjin (March 2025), possibly reflecting localized outbreaks or viral content amplification (see Figures 3-5).

LDA topic modeling, based on perplexity and coherence evaluations (see Figures 6 and 7), determined the optimal number of topics as 8. Topic distance mapping showed minimal overlap (see Figure 8).

To present high-frequency and high-probability associated words under each topic, we ultimately selected 15 practically significant associated words for each topic (see Table 1), it presents the top 15 associated words ranked by P(w|z) for each topic, along with their term frequencies.

We’ve finalized 8 themes based on keyword associations and source text, and labeled them accordingly. For instance, Topic 1 was labeled “Mpox Transmission and Prevention” since its keywords primarily addressed transmission routes and preventive measures. Similarly, Topic 2 was designated “Global Mpox Epidemic Monitoring and Control” as its source text focused on epidemiological surveillance and containment strategies. Corresponding examples for each theme are provided in Table 2.

Topic labels and interpretations are as follows:

To explore public interest during peak search months, we conducted separate LDA topic modeling for the 4 significant peak months: May 2022, July 2023, August 2024, and February 2025. Perplexity and coherence metrics were recalculated for each period to determine the optimal number of topics: 2, 6, 4, and 5 topics, respectively (see Multimedia Appendices 1-8). Topic distances were also examined, and no keyword overlaps were found (see Multimedia Appendices 9-12).

Multimedia Appendix 13 presents the top 15 keywords for each topic during the 4 peak periods of public interest. In May 2022, users primarily focused on acquiring a general understanding of mpox, encompassing transmission routes, prevention methods, and global outbreak status. By July 2023, public attention had shifted toward real-time news updates, clinical features (eg, incubation period), impacts on specific groups (eg, transgender individuals), and newly reported domestic cases. In August 2024, searches reflected increasing concern regarding visual disease manifestations, medical supply availability, transmission mechanism research, and discussions on vaccine efficacy and administration procedures. By February 2025, attention shifted to China’s latest outbreak status, updated prevention strategies, the initiation of mpox vaccine clinical trials, and extensive media coverage. To visualize thematic transitions across these periods, we constructed Sankey diagrams (see Multimedia Appendix 14) illustrating shifts in user interest in response to evolving epidemic dynamics. Line thickness indicates the strength of thematic linkages between time points. For instance, the topic “Mpox Disease Treatment and Prevention Strategies” (July 2023) evolved from “Mpox Transmission and Prevention” and “Global Mpox Epidemic Monitoring and Control” (May 2022). It subsequently transitioned to “Mpox Outbreak Characteristics and Control in China” and “Mpox Symptom Presentation and Transmission Modes” (August 2024), and further diverged into “Mpox Case Detection and Transmission,” “News Updates on Mpox Prevention,” “Mpox Infection and Treatment in China,” and “Mpox Transmission Characteristics” (February 2025). This analysis demonstrates how mpox-related topics evolved from May 2022 to February 2025, shifting from global surveillance and prevention to regional control and practical applications, with an emphasis on case management and vaccine-based prevention within China.

In this study, we analyzed the trends in the mpox-related Search Index on TikTok in China between April 1, 2022, and March 31, 2025. A total of 4 significant peaks were identified in May 2022, July 2023, August 2024, and February 2025. These surges corresponded closely with key global and national public health events. Specifically, in May 2022, the WHO announced the beginning of a global mpox outbreak [31]; in July 2023, the Chinese Center for Disease Control reported 106 newly confirmed mpox cases in June, marking the first update since the initial imported case in Chongqing in 2022 [32]; in August 2024, the WHO declared the mpox outbreak in the Democratic Republic of the Congo a PHEIC [3]; and in January 2025, the clade Ib Mpox was detected in China. Subsequently, in February 2025, multiple major Chinese media outlets reposted the latest WHO outbreak report, drawing significant public attention. These findings suggest that public concern regarding mpox on TikTok is highly sensitive to the progression and reporting of the epidemic.

TGI trend analysis by gender and age revealed important demographic insights. Although monthly gender-based TGI values fluctuated, no consistent gender bias was observed. However, male TGI values remained elevated longer as the outbreak evolved, potentially due to the predominance of male patients in China. Notably, age-based differences were more pronounced: younger users had higher TGI, with the 18‐23 years age group showing the most significant engagement and volatility, peaking in late 2024. The 24‐30 years age group followed, with a more stable pattern. In contrast, TGI among users aged 31‐50 years was relatively low with minimal fluctuation, and those aged ≥51 years exhibited the lowest and most stable TGI values. These differences may be attributed to younger users’ higher engagement with social media platforms and their heightened concern regarding emerging health threats. These findings highlight the need for age-targeted health communication strategies. For instance, older adults, who may have limited social media exposure, might benefit more from offline interventions to improve mpox awareness.

Geographically, provinces such as Beijing, Tianjin, and Jilin exhibited the highest cumulative TGI values, which could reflect better healthcare infrastructure, more efficient information dissemination, and higher public health literacy. During certain months, individual provinces such as Gansu (December 2023) and Tianjin (March 2025) showed sudden TGI spikes. These may have been triggered by local outbreak reports or intensified media coverage. For example, in December 2023, several widely circulated TikTok videos highlighted Gansu in bold text when discussing new mpox cases [33]. Similarly, in March 2025, a viral post featuring a Weibo screenshot of an mpox case in Tianjin’s Jing’an district gained extensive traction on TikTok [34]. These cases suggest that leveraging influential individuals or visual emphasis in digital communication can enhance the visibility and spread of health-related content.

By selecting 8 topics based on perplexity and coherence scores, we were able to comprehensively capture the thematic structure of mpox-related discussions on TikTok. These topics spanned transmission, prevention, surveillance, treatment, vaccination, case reports, and symptomatology—indicating diverse public interests and concerns. However, certain associated keywords reflected ongoing public confusion and challenges to epidemic prevention. The presence of fundamental knowledge gaps was evident. Queries such as “What’s the difference between mpox and chickenpox?,” “Can mpox be cured?,” and “Is mpox still contagious after recovery?” suggest a general lack of understanding. Alarmingly, previous studies have shown that even some sexual health professionals had limited knowledge about mpox [35]. Therefore, enhancing public health literacy and empowering individuals to distinguish accurate from false information are essential to promoting risk awareness and responsible health behaviors.

Misinformation poses a significant challenge in public health communication during the mpox outbreak. Our analysis identified search terms and topics propagating inaccurate or misleading information. For example, some searches suggested that mpox was spread by U.S. planes or developed by American and Jewish capital (eg, “mpox spread by U.S. planes [36],” “virus developed by American and Jewish capital [37]”). Some search terms and topics indicating public mistrust and concerns about mpox vaccines. For instance, questions like “Why is the mpox vaccine not being promoted?” and “Adding HIV to the mpox vaccine” revealed reluctance and skepticism toward vaccination [10,38]. These false claims distort disease etiology and may incite xenophobic sentiment.

While platforms like TikTok facilitate timely information dissemination, they may also host misleading content. A COVID-19 study reported that 27% of TikTok videos in the early phase contained incorrect or incomplete information [39]. In studies on COVID-19 misinformation [38,40], it has been found that its spread is associated with multiple factors. On the one hand, misinformation can provide the public with an explanation of uncertain events, alleviating their anxiety and thus spreading widely. On the other hand, some groups deliberately tolerate or promote its spread for political, economic, or other vested interests. Thus, content regulation by authorities is essential. Health authorities should therefore actively monitor social media content to curb misinformation. For example, to overcome vaccine hesitancy, it is crucial to enhance public health literacy and provide accurate information about the safety and efficacy of mpox vaccines. In addition, it is imperative for the global political, business, and scientific communities to work together and take coordinated actions to preserve the integrity and credibility of professional expertise and rebuild public trust [40,41].

Stigmatization of affected populations has emerged as a notable issue in the mpox-related discussions on TikTok. Our analysis revealed that some queries and statements perpetuated harmful stereotypes, such as the notion that mpox is only transmitted between men who have sex with men or is common solely among homosexuals. Such stigmatizing discourse can marginalize affected groups, leading to self-concealment of cases and reduced vigilance among the general public. This, in turn, may hinder disease control efforts [9,35]. It is essential to address stigma through education and awareness campaigns that promote understanding and empathy toward affected communities.

Public attention also shifted across different stages of the outbreak. In May 2022, when mpox first attracted public attention, searches focused on basic virology and global spread (eg, “What kind of disease is mpox?,” “Mpox symptoms look like smallpox,” “Which countries have outbreaks?”). These trends align with previous findings that individuals seek real-time information during health emergencies via social media platforms [42,43].

As public understanding of mpox improved and the outbreak evolved, public attention shifted [44]. In July 2023, keywords such as “expert interpretation of transmission” and “CDC’s mpox infection risk analysis” reflected the growing presence of professionals in content creation. However, stigmatizing discourse also intensified, including inflammatory posts targeting transgender individuals or referencing celebrities associated with LGBTQ+ communities [39,45] (Some comments have linked transgenders to mpox). In August 2024, following the WHO’s renewed declaration of a PHEIC, public discourse once again shifted, with growing expressions of fear—consistent with previous research showing increased fear-related posts during global health emergencies [46,47]. Keywords such as “photos of mpox pustules” and “disfigurement after recovery” exemplify this trend. In February 2025, the release of WHO’s latest mpox report and wide media dissemination reignited public interest, with a focus on domestic developments such as cluster outbreaks, case statistics, and clinical trials of mpox vaccines in China.

Collectively, mpox-related content evolution on TikTok paralleled epidemic progression and public information needs. This supports TikTok’s utility for real-time health surveillance and underscores the need for dynamic health education. Concurrently, public health efforts must counter misinformation, stigmatization, and fear to ensure effective disease control.

Nevertheless, this study has several limitations. First, technical constraints prevented retrieval of all mpox-related TikTok content and more granular analyses of user subgroups. Second, our data were limited to associated search terms, preventing deeper analysis of the content behind those searches. Third, some internal algorithms (eg, TGI calculation) used by TikTok are not fully transparent, which may affect data accuracy. Fourth, TikTok data may not represent nonusers (eg, older adults with limited digital proficiency), introducing coverage bias. Finally, although comprehensive data cleaning was implemented, context-specific or emerging expressions may have been inadvertently excluded by the filtering system.

This study analyzed TikTok data on mpox in China using LDA topic modeling, revealing that public attention closely followed epidemic developments and varied by age and region. A total of 8 major topics were identified, ranging from transmission and prevention to misinformation and stigmatization. The findings highlight the dual role of social media as both an information source and a potential vector for misinformation. In actual epidemic prevention and control, public health agencies ought to harness TikTok actively. Collaborating with influential bloggers or opinion leaders to devise engaging health information content strategies is a recommended approach, using their sway to steer audiences toward cultivating healthy habits [48]. Meanwhile, establishing a swift monitoring mechanism is essential for detecting misinformation in a timely manner and issuing accurate information to rectify misconceptions [49]. In addition, regularly disseminating authoritative and accurate public health information, maintaining electronic communication with the public to comprehend their needs and concerns, conducting internet-based health education activities, and devising behavioral intervention projects to encourage healthy behavioral changes are also important [50]. Furthermore, analyzing TikTok data to formulate precise communication strategies can elevate the effectiveness of public health communication.