An Exploratory Typology of Tobacco-Related Misleading Content on Social Media: Qualitative Analysis of Instagram and TikTok

Background: Tobacco-related misinformation on social media platforms presents growing challenges to digital health communication and public health. Although prior studies have focused on platform-specific patterns, a unified framework for categorizing and comparing misinformation across platforms is lacking. Such a framework is essential for improving infodemiological surveillance and designing targeted digital interventions. Objective: This study was an exploratory analysis aimed to build a cross-platform typology to categorize tobacco-related misinformation. Methods: Data from Instagram and TikTok between January 2020 and August 2023 were collected using a third-party data collection platform (CrowdTangle) and the TikTok Research application programming interface (API). We reviewed a total of 4850 Instagram posts using a combination of generative artificial intelligence (AI) and human validation by two independent reviewers. In addition, 719 TikTok videos were reviewed manually using qualitative analysis. We iteratively developed and refined the exploratory typology informed by the literature integrating our prior analysis of Twitter data and these new datasets. Results: Of the 22 (71%) Instagram posts and 9 (29%) TikTok videos we analyzed closely to classify misinformation, 2 (6.5%) were about cigarettes, 22 (71%) were about electronic cigarettes (e-cigarettes), 1 (3.2%) was about heated tobacco products (HTPs), 2 (6.5%) were about nicotine (not mentioning specific products), and 3 (9.7%) were about cannabidiol (CBD) products. 1 (3.2%) post did not mention any type of products. These categories could overlap in a single post. The resulting typology consisted of five core narrative archetypes: false or misleading health claims (A1), wellness and lifestyle appeal (A2), conspiracy-driven policy agenda (A3), undermining trust in science and medicine (A4), and recreational nicotine use normalization (A5). Each archetype has attributes of false claim types and sources. Among the posts we analyzed, A1 and A2 were most likely to be found on Instagram. A3 was most frequently found on Twitter. A4 was commonly seen on both Twitter and TikTok, and A5 was most frequently found on TikTok. Two additional dimensions—type of falsehood and source—were also added to characterize a given misinformation post. This exploratory typology paved the way for a structured lens to view how misinformation is tailored to digital environments and target audiences. Conclusions: This cross-platform typology building supports digital health research by integrating AI and qualitative methods to categorize tobacco-related misinformation. It can inform the development of automated misinformation detection models, enhance real-time infodemiological monitoring, and guide digital public health campaigns to build tailored countermessaging.