Four hundred and thirty participants aged 18-24 years in the United States completed the 15-minute study via Amazon Mechanical Turk [28]. The study was approved by the Institutional Review Board of the University of Pennsylvania. Participants were compensated US $1 for study completion. Researchers have found psychometric indicators of quality of Mechanical Turk data to be comparable to subject pools at research universities [29]. Nevertheless, we included two checks on participation quality [30]. To ensure that participants read the instructions, they were required to answer a question about their favorite color by clicking green and pink regardless of their actual preferences. Failing to follow this instruction indicates that this, and possibly other instructions of questions, were not read. A second check ensured that participants actually watched the videos by indicating what was discussed. Participants responded to a checklist of products including the tobacco product that appeared in the video and indicated which products appeared or were mentioned. Participants failing either check (80/430, 18.6%) were excluded, producing a final sample size of 350. Comparisons between the excluded and retained participants indicated no significant differences in tobacco consumption or demographics.

A search of popular videos on YouTube using 136 tobacco-related search terms identified over 8000 videos after removing unrelated content. Search terms included “smoke,” “smoking,” “tobacco,” “cigarillo,” and colloquial terms for products such as “shisha” for hookah tobacco. Criteria for exclusion included the video having fewer than 20,000 views at the time of download, the video being in a language other than English or having no audio, a video not containing tobacco content, or a video not being retrievable due to a broken or inactive ink. Using an Excel random number generator, of the 8000 eligible videos, 200 were selected for further coding. Three coders met a Krippendorff alpha reliability of K alpha >.91 for the classifications of videos into different types of claims. This coding identified four major types of misleading health messages from this sample of YouTube videos: (1) rejection of science (ie, evidence supporting the harmfulness of a tobacco product is faulty), (2) assertion of benefit (eg, tobacco can be healthy), (3) denial of harm (eg, tobacco is not harmful after all), or (4) presence of acceptable risks (eg, using tobacco is no riskier than other common activities).

We pinpointed 37 videos containing misleading portrayals of tobacco’s health consequences that lacked discernable brand affiliation or sponsorship. To ensure a varied sample of contents, we selected four videos representing each of the misleading categories. Within this selected set, two of the videos featured young adult white males and two featured adult white males. Within the selected set, videos were also representative of the major categories of misleading health claims that reached large audiences. In the selected videos, the source: (1) claimed that drinking green tea prevents mouth disease from chewing tobacco (denial of harm), (2) expressed skepticism toward scientific evidence that shisha contains harmful additives and that water filtration does not eliminate carcinogens (rejection of science), (3) suggested routine tasks like driving a car entail risks similar to pipe smoking (relative risk), or (4) asserted his status as a fitness expert while vaping (assertion of benefit). We selected videos that were popular without presenting expert sources. This selection allowed for the most stringent test of possible consequences of seemingly harmless amateur videos posted on YouTube. In addition to the lack of connection to a brand or sponsor, this selection also ensured a low probability of capturing commercial content.

The control video was not related to health. This YouTube video featured a demonstration about replacing a shower faucet and was similar in duration and features to the experimental videos. Videos were cropped to minimize background and edited into brief (approximately 20 seconds) segments. Screenshots and links to the videos appear in Figure 1.

The conditions included five videos–four experimental videos and the control video– randomized between participants. Attitudes were measured after participants viewed one of the five videos. Specifically, in the experimental video conditions, we measured attitudes towards (1) chewing tobacco, (2) e-cigarettes, (3) hookahs, and (4) pipe smoking, depending on which was featured in the video. In the control video condition, we measured attitudes towards (1) chewing tobacco, (2) e-cigarettes, (3) hookahs, and (4) pipe smoking in all cases. Attitudes toward combustive cigarette smoking were measured in all conditions even though combustive cigarettes were not featured.

For descriptive purposes, we assessed use for the following products: cigarettes, pipe tobacco, chewing tobacco, hookahs, and e-cigarettes. We specifically asked participants, “Have you ever used any of the following tobacco products?”, after which they checked products that they had used. Table 1 contains the demographic and tobacco product usage description of the sample. The sample had considerable experience smoking combustive cigarettes, using hookahs, and using e-cigarettes but had little experience chewing tobacco and smoking pipes. As shown by the inferential statistics used to compare across experimental groups, there were no significant differences across the five video conditions in any of these characteristics. These analyses thus suggested that any experimental effects were due to the videos rather than a priori differences among groups of participants.

The level of credibility of the spokesperson in the experimental videos was low to moderate, as judged by a mean credibility of 2.52 (SD 0.56), which differed significantly from the midpoint (3) of the scale, t (279)=–14.24, P<.001. The means and 95% confidence intervals were 2.30 (2.18-2.43) for chewing tobacco, 2.50 (2.38-2.62) for hookahs, 2.88 (2.76-3.01) for pipes, and 2.41 (2.29-2.53) for e-cigarettes; all suggestive of the low to moderate credibility of amateur sources.

We estimated attitudes toward featured products across each specific experimental video and the control condition. As explained above, each experimental condition measured attitudes toward a different featured product, whereas the control condition measured attitudes towards all products, which were averaged for comparison. Table 2 presents these analyses; post hoc least significant difference (LSD) contrasts are represented with different subscripts. As shown, attitudes toward e-cigarettes and hookahs were more favorable following the experimental videos compared to the control video. In addition, the video about chewing tobacco produced more negative attitudes toward chewing tobacco that did the control video, and the video about pipe smoking did not differ from the control.

The effects on attitudes were investigated by conducting analyses of variance of attitudes toward the use of combustive cigarettes (never featured in the presented videos) as a function of experimental condition. Results from these analyses appear in the lower panel of Table 2 and show a significant effect of condition for both overall attitudes and positively worded attitude items. The omminbus effects on these attitude measures can be attributed to significant differences between the chewing and pipe videos, between the chewing and the e-cigarette videos, and between the hookah and the pipe videos. Although none of the videos differed significantly from the control videos, the pipe video produced the most favorable attitudes toward combustive cigarette smoking. Specifically, the pipe video led to more positive attitudes toward combustive smoking than did the chewing and the hookah videos, and the e-cigarette video led to more positive attitudes toward combustive cigarette smoking than did the chewing video.

We examined the responses of young adults aged 18-24 years realted to four misleading portrayals of tobacco’s health consequences in popular YouTube videos. Results indicated that such material can increase positivity toward the featured products, such as e-cigarettes or hookah smoking (standardized difference Cohen g vs control video in each case was 0.38 and 0.37).

Although our study illustrates how potentially harmful content on social media may be studied, it has some limitations. We cannot generalize our findings to the many other videos that populate YouTube. We attempted to cover the major types of claims made in those videos, but there may well be others that are even more persuasive than the ones we identified. In addition, our findings with young adults may not generalize to adolescents, who may be even more susceptible to the claims made in these videos. Further research will be needed to assess this possibility.

This tendency for young tobacco consumers to respond credulously to misinformation on YouTube raises the possibility of tobacco use and exposure to misleading media exerting mutually reinforcing effects [31]. Finding tobacco-friendly material convincing and gratifying, recipients might seek out similar content, further bolstering self-justifying beliefs and prompting further selective exposure. In this regard, the aforementioned abundance of material on YouTube is cause for concern.

The present findings highlight the need to further study how new media sources such as YouTube affect tobacco knowledge, attitudes, and behavior. Although this study investigated misleading health portrayals only, content might also affect perceptions of tobacco products in other ways, both blatant and subtle, such as by modeling consumption [1-3], associating products with sex [16], or facilitating product acquisition [20]. Furthermore, the participatory, interactive, and self-selected nature of social media may enhance pro-tobacco media effects more readily and perniciously than was possible with traditional media.

The findings of YouTube video effects on positive attitudes toward hookahs and e-cigarettes should alert the public to the potential threats that these widely viewed videos can pose to youth and the health of the population. Attitudes toward hookahs and e-cigarettes can predict engagement in the behavior in the future [32-35], and essentially mimic the long-standing strategy of the tobacco industry to create favorable impressions of their products despite the harm they cause [36,37].