Another post not on a Thursday? Yes, friends, I am writing a day early because today two reports came out that I think are important for us to see. I’ll get back to regular cadence next week!
For now, enjoy and let me know your thoughts (and definitely read the reports!).
The next generation, the one growing up with technology unlike anything any other generation has seen, is going to shape how the world uses these tools.
I’m sure your world, like mine, has a lot of opinions about technology, especially AI, and what it means for mental health, jobs, and everything else. As a colleague said recently, the current version of AI we’re using is going to be the least refined, because it gets better every day.
But the youth, the ones having these technologies rain down on them in unprecedented ways, what do they think? For example, what if you don’t like AI giving you a summary instead of letting you find the best place yourself?
I sometimes talk about AI like it’s a distant cousin, or a friend who shows up uninvited. Whatever your thoughts on AI, the truth is it’s already in the room; in fact, it skipped ringing the doorbell and made itself comfortable on the couch all without you really knowing. So, now we do the boring adult part, we acknowledge it, set some boundaries, and use it responsibly, like we would any other tool.
And this isn’t niche. ChatGPT is used at huge scale, with the U.S. and India leading the pack.
That’s why these new reports matter.
Today, two complementary reports dropped that try to answer the question everyone keeps arguing about while rarely asking young people directly. The first looks at how youth are using GenAI. The second zooms in on youth mental health and AI. The first report was done in partnership with Young Futures, and the second was done in partnership with The Jed Foundation (JED). Both are worth a minute of your time to review.
They’re part of the Youth Mental Health Tracker (YMHT), created by Surgo Health (a Public Benefit Corporation). The research is based on a nationally representative survey of more than 1,300 youth ages 13 to 24.
There’s a lot here, and it’s useful for how we talk about youth, mental health, and AI.
Youth and GenAI: Why Context Matters
The first report pushes back on a lazy assumption I’ve been seeing that’s driving a lot of the debate, that the youth AI experience can be measured by time spent or frequency. Frankly, simply looking at just use is too blunt a metric to be useful as frequency alone is likely a poor indicator of risk or benefit.
What is apparent is that young people engage with AI in profoundly different ways, and those differences track with real life. Stress. Discrimination. Access to support. Opportunity, or the lack of it. They all show up with AI too.
The report describes six patterns of use, each with its own mix of upsides and risks. Low-use anxious skeptics. Thriving light-touch pragmatists. Worried strivers. Emotionally entangled superusers. High-hope, high-use skill builders. Curious low-concern learners. The biggest group is the Thriving Light-Touch Pragmatists at 32 percent.
When you look at the report, it’s clear that both low and high users show up across these six groups. So “more AI” or “less AI” doesn’t automatically tell you whether someone is doing okay. The report also finds that historical inequities don’t disappear just because the interface looks new or clean or fancy. They show up in how young people relate to AI.
Black youth are four times more likely than White youth to be “high-hope users,” using AI for learning, creativity, and future-building. Hispanic youth are three times more likely than White youth to be anxious or intense users, leaning on it for coping, stress, and isolation. LGBTQ youth are more than twice as likely to be safety-concerned skeptics, limiting use not because they can’t, but because they’re worried about surveillance, safety, and losing control.
Youth Mental Health in the AI Era: How AI Enters Help-Seeking Pathways
The second report narrows in on a smaller slice of the sample. It looks at the 12 percent of youth who say they’re experiencing mental health struggles and used GenAI during periods of distress.
Even in this report, there’s not just one story. Some youth use AI alongside professional care, or before and after it. Others use it as their only support, a substitute, not a supplement. I tried to make a table to summarize some of these survey data, including why it matters for us.
What’s hard to ignore is where these conversations are happening. Most youth in this group are relying on general-purpose AI tools, not systems designed or safeguarded for mental health support. So disclosures are landing on products that weren’t built for this job, during the moments when the stakes are highest.
In the report, AI shows up in three roles: a tool for coping in the moment, a bridge toward care, or a substitute for care. Those are not the same thing, and mixing them together is how we end up arguing past the actual risk.
When help takes weeks, anyone will grab what’s available tonight. That’s why the “substitute” finding matters so much, it’s less about preference and more about access.
And it’s not all upside. A meaningful share of youth later wish they hadn’t shared something hard or personal with an AI tool. Relief in the moment doesn’t always translate into something that feels good afterward.
The difference between AI as a bridge and AI as an endpoint is the whole safety question. If a tool becomes the default listener for kids with the fewest options, we should treat that as a public health design problem, not just some quirky trend.
And for about one in three, AI isn’t a bridge at all, it’s the substitute for professional care. While that feels like a lot, it’s also a pretty stark reminder that these tools are getting used not because they all love chatbots, but because barriers like insurance coverage, transportation, or broader access block the usual paths.
When I read these reports, something kept nagging at me.
I couldn’t help but be reminded of how our system is continuing to fail young people. Safety. Immigration. Racism. Economics. Oh and that other perennial issue, accessing health care, which remains too expensive, slow, confusing to access, and overly dependent on whether an adult can help you navigate it. When help takes weeks, anyone will grab what’s available tonight. So as we debate the tool, we should look hard at the gap it’s being asked to fill.
So what do we do?
First, we have to bring young people into conversations around AI. Not as a token panel, not just the “youth perspective” with smiling photos on a slide deck at the beginning or end, but as co-designers from the start. Ask young people what feels useful, what feels creepy, and what makes things worse. Then we should treat those answers like inputs, not just decorative call out quotes.
Second, stop treating youth who use AI as one big group. The first report makes it clear that a low-use kid can be anxious and skeptical, and a high-use kid can be thriving and building skills, or struggling and emotionally tangled. Once again, if we care about building effective programs, we can’t treat all people, all groups, the same. Sure, it’s the same tool, but considering the wildly different relationships we have with it, our response can’t be one-size-fits-all either.
We also need to be honest about tradeoffs. If a tool helps one group learn faster but makes another group feel watched, that’s an important story.
We also need to start where the decisions actually get made. Schools deciding what’s allowed and what’s not should have youth in the room, not just adults guessing what “students will do anyway.” Platforms writing policies should stop pretending everyone has the same level of safety, privacy, and support. They don’t!
From there, build a few practical structures that make this real. Youth advisory groups with actual power, not just a photo op. Student-led pilots where young people help test what belongs in classrooms and what doesn’t.
Finally, clear privacy rules written in plain English would help everyone. If it takes a lawyer to translate, it likely won’t protect kids in real life. We need protections that match the because the kids leaning on AI the most are often the ones carrying the most already. We knew that, or should have known that, ever before AI dominated the news cycle.
If we don’t include youth now, we’ll build a future ‘for’ them that they never asked for, and they’ll still be the ones stuck living in it.
This post is 100 just on the volume of analysis shared, and those amazing links that IDK (Youth Mental Health Tracker!!!!). I'm so avoiding doing scheduled work to deep dive into it all.
I was grabbed by the short hairs because just the other night, doing my Crisis Intervention Volunteering, a teen noted that I sound like Google AI while being inconsistent with my grammar skills. (I own that, since paying attention in grade school was optional, I missed grammar.) They also noted that I was not nagging them to go to 988 when they talked about self-harm. They were there.
What made me scratch my head after I got this teen to safety? In fact, they said, "I feel so normal now, thank you." I wondered exactly about the entire examination of teens & AI you present and the links shared. I wondered... Is AI slowing or curbing teens going to sites like 988 or CTL? Are teens coming into those sites better prepared to chat, share, and learn? Is AI accelerating teens seeking support on these sites or professional help? It is there.
Ben, I think you captured that well: "And for about one in three, AI isn’t a bridge at all, it’s the substitute for professional care. While that feels like a lot, it’s also a pretty stark reminder that these tools are getting used not because they all love chatbots, but because barriers like insurance coverage, transportation, or broader access block the usual paths."
Your ending this with practical structures is perfect. Still, I want to know when teens engage in AI for mental health and seek support, what do they bring with them from AI? I guess it's all there; I need to dive into all the reports.
You said it best: "Sure, it’s the same tool, but considering the wildly different relationships we have with it, our response can’t be one-size-fits-all either." Indeed, all of this needs to be choreographed as well as Swan Lake. Not sure a single person like Tchaikosky is able to do. This is teams and policy.
“Youth and GenAI: Why Context Matters”
You’re pointing to something essential: an AI can only be genuinely helpful when it understands the individual in front of it — their context, their constraints, their agency. But today’s systems still don’t know how to engage people in a way that actually learns that context. They default to being gentle — and quietly addictive — know it alls rather than partners.
The real value comes when AI works like a wingman: bounded, collaborative, and aligned with the user’s goals. That’s why your point about not treating youth as one big group is so important. Their relationships with AI reflect their lived realities — stress, identity, support, opportunity — and the tech world has barely begun to design for that.
If we don’t build AI that can adapt to those differences, we’ll keep missing the kids (and adults) who need the most support. And that’s where policy leadership still matters — not by trying to engineer or legislate the solution, but by making it impossible for the tech world to ignore the fact that AI must meet people as individuals, not abstractions.