One in three young adults in a US survey had asked AI about substance use

Generative AI has become one place some young adults ask questions about alcohol and other substances. In a US online survey,321 of 995 respondents aged 18–29 said they had used it for substance-related information. The finding, published in JMIR Formative Research on 19 October, identifies an information-seeking behaviour that health researchers are beginning to measure.[1]

The share was 32.3%, with a reported 95% confidence interval of 29.3%–35.2%. Participants were recruited through the Prolific online research platform. Although the study used demographic quotas, it was a non-probability sample, so the headline finding should remain attached to the surveyed group rather than presented as a definitive rate for every US young adult.[1]

What people were asking

The prominence of consequences suggests that many people were seeking information about effects and risks, rather than only identifying a substance. The survey does not reveal whether the questions were about respondents themselves, someone else or general curiosity in every case. Asking about a disorder is also different from receiving a diagnosis.[1]

Generative AI sat alongside established search habits. Google was used for substance-related information by 572 respondents overall, or 57.5%. Among the 321 AI users,298—92.8%—also reported Google use. The overlap suggests an additional route to information rather than a clean replacement of conventional search.[1]

A substantial group used it repeatedly

Frequency was divided. Of the 321 AI users,164 said they hardly ever used it for this purpose. Another 155, or 48.2%, reported at least monthly use:103 monthly,39 weekly and 13 daily. Two selected ‘never’ in the frequency item despite reporting prior use. Nearly half of the AI-using group therefore described a recurring habit, not merely an isolated experiment.[1]

That matters because an answer can influence what someone searches for next, which risks they notice or whether they seek additional support. The survey establishes that people report using this channel; it does not test the responses they received, whether they acted on them or whether the interaction improved their health.[1]

An information channel that needs its own evidence

A chatbot can make a sensitive question easier to phrase and can offer an immediate explanation. The same conversational format can make uncertainty hard to see if a model supplies an assured answer without the right evidence or context. Those features help explain why researchers need to study the quality of the channel separately from its popularity.[1]

The study's design also has the familiar limits of self-report. Participants had to recall their use and interpret the survey's categories. Recruitment through an online platform may select people more comfortable with digital tools than the wider population. The results provide an empirical starting point, with representative follow-up and direct answer evaluation needed for different questions.[1]

For public-health communication, the immediate implication is reach. A substantial share of this sample had already taken substance-related questions to AI, often while also using search engines. Understanding what they ask, how frequently they return and how they judge the response is becoming part of understanding where young adults find health information.[1]

Report an error or suggest a correction