We are raising adolescents inside an attention economy that does not simply reflect culture; it helps manufacture the culture they experience as normal. Cannabis, nicotine, alcohol, beauty, status, anxiety, humor, and belonging arrive in the same continuous stream, with few reliable boundaries between advertising and friendship, or entertainment and influence. This study does not prove that Instagram, Snapchat, or TikTok causes substance use. It does show that a teenager’s digital setting is not neutral, and that the platforms occupying the most intimate hours of adolescence may travel with different patterns of risk.

What this says about where we are is less comfortable than a demand to delete one app. We have built a culture in which young people are expected to manage industrial-scale persuasion with still-developing judgment, while adults debate screen time as though minutes alone explain the problem. The better response is not panic, blame, or surveillance. It is to take the environment seriously: ask what adolescents are seeing, whose approval matters, how cannabis is being framed, and whether families, schools, clinicians, platforms, and regulators are willing to share responsibility for the norms being produced.

Among adolescents without cannabis use at the analytical baseline, each additional daily hour of Instagram was associated with 49% higher odds of newly reported cannabis use one year later. The confidence interval ranged from 9% to 98%, and the result remained significant after false-discovery-rate correction.

YouTube time showed an inverse association with newly reported cannabis use. These estimates describe patterns within this cohort and do not establish that increasing or decreasing time on either platform would change an individual teenager’s risk.

At the analytical baseline, cannabis use was positively associated with time on TikTok, Instagram, and Snapchat after adjustment for age, sex, parental education, race or ethnicity, and overall phone use. YouTube time was inversely associated with cannabis use.

Alcohol and nicotine or tobacco associations also differed by app. The pattern argues against treating all social media exposure as interchangeable, but cross-sectional associations cannot determine whether app use preceded substance use.

Among adolescents who already reported substance use, the prospective models found few durable relationships between app hours and later use days, cannabis-related problems, or nicotine craving.

Several nominal associations disappeared after false-discovery-rate correction. Those null and corrected-away results are important because they limit any claim that more time on a particular app consistently predicts worsening substance use.

The EARS system recorded foreground app use directly for three weeks, reducing the recall error that affects self-reported screen-time estimates. Researchers could compare actual time on six common applications while adjusting for overall phone use.

The sensing data did not reveal which posts, advertisements, private messages, or videos participants saw. Time on an app is therefore not the same exposure as pro-cannabis content, peer approval, marketing, or algorithmic recommendation.

Only ABCD participants with personal Android phones could enter the sensing substudy. In ABCD, Android users were more likely than iPhone users to be male and from lower-income families, and 59% of this analytical sample was male.

Substance-use prevalence was low and prospective initiation events were few. Replication with Apple sensing data, longer follow-up, and more diverse patterns of use is needed before treating the estimates as broadly generalizable.

The findings fit a larger literature linking social-media exposure, peer norms, and positive substance portrayals with adolescent risk. This paper adds platform-specific passive sensing, but it does not directly measure the messages or social interactions that might carry that influence.

Prevention may be more useful when it combines media literacy, family communication, confidential clinical screening, and attention to peer context rather than relying on blanket screen-time rules. Those approaches remain sensible even while causal evidence about individual platforms develops.

The strongest clinical contribution is not a ranking of good and bad apps. It is the reminder that digital environments differ, and that time alone is an incomplete proxy for what a young person sees, shares, and learns from peers.

I would use this study to open a nonjudgmental conversation about cannabis exposure, peer norms, online content, and early use. I would not use it to blame a teenager, confiscate a phone as a medical treatment, or claim that one platform caused the behavior.

A Four-Step Reading Frame