We collected up to 100 videos posted on or before July 1, 2022, from each of the top 100 influencers on TikTok in the United States. Each video was collected by capturing the full video and captions on a research assistant’s phone. Research assistants noted the date the video was posted and the numbers of views, likes, comments, and “saves” or bookmarks. They also collected profile information, such as whether the influencer was “verified” with a blue checkmark and the total number of likes and followers.

This study was determined to be non–human subjects research and used preexisting data that included publicly available information and therefore does not require institutional review board review per Federal Regulations for the Protection of Human Research Subjects (45 CFR 46.104(d)(4)) [22]. TikTok allows users to select whether their accounts are public or private. All collected data (ie, videos) were publicly available to anyone on TikTok. No user (ie, follower) data were collected. Only deidentified and aggregated data are presented.

Using a 3-step approach, we first identified whether a food product was featured. Second, we identified the main product, based on whether it was (1) the only featured food product; (2) a sponsored product; or (3) the most prominent or dominant product (eg, the first to appear or to grab the viewer’s eye). For each main product, the research team then noted whether it was branded (eg, Coke bottle vs banana). Third, a team of research assistants viewed each food product video to capture information about how the product was featured (ie, in the video, caption, or audio) and, for any branded product, where the brand or logo was featured and whether the video was accompanied by any brand disclosures.

Using ReCal3 [23], we calculated interrater reliability among coders in the third step using a small sample of “practice” videos (n=6) that reflected the range of disclosures used [24]. Average pairwise percentage agreement was 92%, and average pairwise Cohen κ was 0.82. We deemed these to be acceptable given the descriptive and exploratory nature of this work. We report descriptive statistics.

Nearly half of the TikTok posts by top influencers that included food and beverage products featured a branded product. The majority (69.3%) of these posts, however, did not include any brand relationship disclosures, suggesting either that no relationship existed or that they purposely or accidentally omitted its disclosure. By comparison, a recent study of food-related Instagram posts by celebrities (athletes, actors, and musicians) found that less than 5% of the identified posts that were sponsored by a food company were identified by hashtags (eg, “#ad”) or related terms (eg, “paid advertisement”) below the post [25].

Among the food videos with disclosures we identified, the most frequently used mechanism—tagging the brand—did not clearly differentiate between sponsored content and the influencer trying to attract a brand or followers who may like that brand. These findings have implications for how social media marketing should be regulated. The ambiguity around brand relationship disclosures makes it difficult for social media users, particularly adolescents, to make informed decisions about the products they see [18]. This concern has been recognized by the FTC in the most recent iteration (July 2023) of its endorsement guides [26], which emphasize that marketing practices geared toward adults may not be appropriate for children. In fact, bigger and clearer disclosures would likely work better among children and adolescents. A qualitative study among participants aged 12-16 years reported that they did not mind if influencers were paid, but they did not like seeing explicit disclosures that were “...too pushy, like, this is advertising, because then you will keep this in mind all the time” [27].

Given the dearth of information on influencer marketing, our study set out to systematically collect and record information on how social media influencers disclose relationships with food brands. However, our study has limitations, including (1) the use of a convenience sample of influencers and posts, (2) the focus on only the most prominent food product in each video, (3) the inability to infer intentionality in whether a product was meant to be featured in a video (eg, when the only food product featured was in the background), and (4) the inability to distinguish whether the presence or lack of a disclosure implied the presence or lack of a brand relationship.

Social media users, particularly adolescents, need clearer, more robust disclosures from influencers to protect against the undue influence of food marketing. These findings may inform calls for the Children’s Food and Beverage Advertising Initiative—the largest self-regulatory pledge designed to reduce unhealthy food marketing—to extend its protections to older adolescents, who are heavily targeted by food companies [28,29].