How Reddit, TikTok, and AI are changing the game for substance use researchers

When you think of tools for studying substance use and addiction, a social media site like Reddit, TikTok, or YouTube probably isn’t the first thing that comes to mind. Yet the stories shared on social media platforms are offering unprecedented insights into the world of substance use.

In the past, researchers studying peoples’ experiences with addiction relied mostly on clinical observations and self-reported surveys. But only about 5% of people diagnosed with a substance use disorder seek formal treatment. They are only a small sliver of the population who have a substance use disorder—and until recently, there has been no straightforward way to capture the experiences of the other 95%.

Today, millions of people openly discuss their experiences with drugs online, creating a vast collection of raw narratives about drug use. As a doctoral student in information science with a background in public health, I use this material to better understand how people who use drugs describe their lives and make sense of their experiences, especially when it comes to stigma.

These online conversations are reshaping how researchers think about substance use, addiction, and recovery. Advances in artificial intelligence are helping make sense of these conversations at a scale that wasn’t possible before.

The hidden population

The vast majority of people diagnosed with a substance use disorder address the issue informally—seeking support from their community, friends, or family, self-medicating or doing nothing at all. But some choose to post about their drug use in dedicated online communities, such as group forums, often with a level of candor that would be difficult to capture in clinical interviews.

Their social media posts offer a window into real-time, unscripted conversations about substance use. For example, Reddit, which is comprised of topical communities called subreddits, contains over 150 interconnected communities dedicated to various aspects of substance use.

In 2024, my colleagues and I analyzed how participants in drug-related forums on Reddit connect and interact. We found that they focused on the chemistry and pharmacology of substances, support for drug users, recreational experiences such as festivals and book clubs, recovery help, and harm reduction strategies. We then selected a few of the most active communities to develop a system for categorizing different types of personal disclosures by labeling 500 Reddit posts.

Policymakers and public health experts have expressed concerns that social media encourages risky drug use. Our work did not assess that issue, but it did support the notion that platforms such as Reddit and TikTok often serve as a lifeline for people seeking just-in-time support when they need it most.

When we used machine learning to analyze an additional 1,000 posts, we found that most users in the forums we focused on were seeking practical safety information. Posters often posed questions such as how much of a substance is safe to take, what interactions to avoid and how to recognize signs of trouble.

We observed that these forums function as informal harm reduction spaces. People share not just experiences but warnings, safety protocols and genuine care for each other’s well-being. When community members are lost to overdose, the responses reveal deep grief and renewed commitments to keeping others safe. This is the everyday reality of how people navigate substance use outside medical settings—with far more nuance and mutual support than critics might expect.

We also explored TikTok, analyzing more than 350 videos from substance-related communities. Recovery advocacy content was the most common, depicted in 33.9% of the videos we analyzed. Just 6.5% of the videos showed active drug use. On Reddit, we frequently saw people emphasizing safety and care.

Why AI is a game-changer

Platforms like Reddit, TikTok, and YouTube host millions of posts, videos, and comments, many filled with slang, sarcasm, regional language, or emotionally charged stories. Analyzing this content manually is time-consuming, inconsistent, and virtually impossible to do at scale.

That’s where AI comes in. Traditional machine learning approaches often rely on fixed word lists or keyword matching, which can miss important contextual cues. In contrast, newer models—especially large language models like OpenAI’s GPT-5—are capable of understanding nuance, tone, and even the underlying intent of a message. This makes them especially useful for studying complex issues like drug use or stigma, where people often communicate through implication, coded language, or emotional nuance rather than direct statements.

These models can identify patterns across thousands of posts and flag emerging trends. For example, researchers used them to detect shifts in how Canadians on X, the social media site formerly called Twitter, discussed cannabis as legalization approached—capturing shifts in public attitudes that traditional surveys might have missed.

In another study, researchers found that monitoring Reddit discussions can help predict opioid-related overdose rates. Official government data, like that from the Centers for Disease Control and Prevention, typically lags by at least six months. But adding near-real-time Reddit data to forecasting models significantly improved their ability to predict overdose deaths—potentially helping public health officials respond faster to emerging crises.

The role that stigma plays in substance use disorder is difficult to capture in traditional surveys and interviews.

Bringing stigma into focus

One of the most difficult aspects of substance use to study—and to address—is the stigma.

It’s deeply personal, often invisible and shaped by a person’s identity, relationships, and environment. Researchers have long recognized that stigma, especially when internalized, can erode self-worth, worsen mental health, and prevent people from seeking help. But it’s notoriously hard to capture using traditional research methods.

Most clinical studies rely on surveys or interviews conducted at regular intervals. While useful, these snapshots can miss how stigma unfolds in everyday life. Stigma scholars have emphasized that understanding its full impact requires paying attention to how people talk about themselves and their experiences over time.

On social media platforms, people often discuss stigma organically, in their own words and in the context of their lived experiences. They might describe being judged by a health care provider, express shame about their own substance use or reflect on how stigma shapes their relationships. Even when posts aren’t directly naming the experience as stigma, they still reveal how stigma is internalized, challenged or reinforced.

Using large language models, researchers can begin to track these patterns at scale, identifying linguistic signals like shame, guilt or expressions of hopelessness. In recent work, my colleagues and I showed that stigma expressed on Reddit aligns closely with long-standing stigma theory—suggesting that what people share on social media reflects recognizable stigma processes, not something fundamentally new or separate from what researchers have long studied.

That matters because stigma is one of the most significant barriers to treatment for people with substance use disorder. Understanding how people who use drugs talk about stigma, harm, recovery, and survival, in their own words, can complement surveys and clinical studies and help inform better public health responses.

By taking these everyday expressions seriously, researchers, clinicians and policymakers can begin to respond to substance use as it is actually lived — messy, evolving and deeply human.

Layla Bouzoubaa is a doctoral student in information science at Drexel University.

This article is republished from The Conversation under a Creative Commons license. Read the original article.