Alternative Health and Conventional Medicine Discourse About Cancer on TikTok: Computer Vision Analysis of TikTok Videos

Background: Health misinformation is abundant online and becoming an increasingly pressing concern for both oncology practitioners and patients with cancer. On social media platforms, including the popular audiovisual app TikTok, the flourishing alternative health industry is further contributing to the spread of misleading and often harmful information, endangering patients' health and outcomes and sowing distrust of the medical community. The prevalence of false and potentially dangerous treatments on a platform that is used as a quasi-search engine by young people poses a serious risk to the health of patients with cancer. Objective: This study seeks to examine how cancer discourse on TikTok differs between alternative health and conventional medicine videos. It aims to look beyond mere facts and falsehoods that TikTok users may utter to understand the visual language and format used in the support of both misleading and truthful narratives, as well as other messages. Methods: Using computer vision analysis and subsequent qualitative close reading of 831 TikTok videos, this study examined how alternative health and conventional medicine videos on cancer differ with regard to the visual language used. Videos were examined for the length of time and prominence in which faces are displayed, as well as for the background setting, location, and Results: The results show that the alt-health and conventional health samples made different use of the audiovisual affordances of TikTok. First, videos from the alternative health sample were more likely to contain a single face that was prominently featured (making up at least 7.5% of the image) for a substantial period of time (35% of the shots), with these testimonial-style videos making up 28.5% (93/326) of the sample compared to 18.6% (94/505) of the conventional medicine sample. Alternative health videos predominantly featured cool tones (P<.001) and were significantly more likely to be filmed outdoors (P<.001), whereas conventional medicine videos were more likely to be shot indoors and feature warm tones such as red, orange, or yellow. Conclusions: The findings of this study contribute to an increased understanding of misinformation as not merely a matter of individual falsehoods but also a phenomenon whose effects might be transported through emotive as well as rational means. They also point to influencer practices and style being an important contributing factor in the declining health of the information environment around cancer and its treatment. The results suggest that public health efforts must extend beyond correcting false statements by injecting factual information into the online cancer discourse and look toward incorporating both visual and rational strategies.