There are two stories I’ve been thinking about and tracking: The recent Common Sense Media survey on teenagers and AI chatbots and a law in Illinois that effectively bans AI therapists.
First, the survey (full report here). It is one of the more helpful snapshots we’ve had of this rapidly evolving relationship between teens and AI. If you care about technology, youth culture, or the policy debates that follow, it’s worth reading in full.
And yet, while the headline from Common Sense Media’s tech guide on teens and AI chatbots is alarmist—calling chatbots an “unacceptable risk” to teens—the data of their survey is more nuanced, and in some cases even encouraging.
What the Numbers Actually Say
As for why teens are using chatbots, the survey let respondents choose up to three reasons. The top two?
- Entertainment (30%)
- Curiosity about the technology (28%)
That’s hardly the dystopian vision of lonely teens pouring their hearts out to synthetic companions. Yes, there’s cause for vigilance—but the dominant motivations look a lot like the early days of any new tech: novelty and fun.
Two other findings jumped out at me:
- 67% of teens say AI conversations are less satisfying than talking with a human.This is a key data point if you’re concerned about kids replacing real relationships with digital ones. The majority of teens are telling us that AI chats don’t measure up. 21% said “about the same.” On the one hand, this can sound depressing. But think about how you might answer the question the survey posed: “How do conversations with AI companions compare to conversations with your real-life friends?” You might think: “All conversations with my friends are always more interesting than AI!” (Certainly your friends will hope so). But not everyone has the same quality or number of friends. How teens answer this question perhaps says more about their friendships than AI.
- 80% spend more time with friends than with AI (and another 13% split their time evenly).In a world where every emerging tech trend gets framed as an impending social apocalypse, these numbers should be a relief. Most teens are still spending most of their time with other people.
A Case for Measured Concern, Not Panic
Commentators like Mandy McLean see this data and focus on the risks. I see reasons for cautious optimism—reasons that should temper the rush to heavy-handed policy.
First, these results are just a snapshot. We don’t have longitudinal data showing whether chatbot use is increasing at the expense of human relationships. The real test will come with time. If these numbers shift dramatically in six months or a year, we’ll know we have a bigger problem.
Second, the proposed “solutions” being floated could create more harm than good. McLean, for example, recommends age-gating AI companions, pointing to the UK’s new “harmful content” framework as a model. But that policy is already showing signs of overreach and unintended consequences—especially in restricting lawful speech and blocking beneficial uses of AI.
Where Bad Policy Is Already Happening: Illinois vs. Utah
We don’t have to look across the Atlantic to see how reactive policymaking can backfire. Illinois recently passed the Wellness and Oversight for Psychological Resources Act, which effectively bans AI in therapy settings.
The law limits AI to “administrative support” or “supplementary support” in therapy or psychotherapy services—and only where a licensed professional takes full responsibility for all interactions. Even then, the scope is narrow. This isn’t “guardrails”; it’s a near lockout.
Contrast that with Utah’s more flexible approach–which I’ve praised for its sensible approach. Utah acknowledges the potential risks of AI therapy but allows for use within a specific context rather than outright prohibition. Their framework leaves room for experimentation, innovation, and the possibility that AI could actually improve access to mental health support—something in critically short supply.
Now we have a natural policy experiment playing out in real time:
- Illinois: Maximal restrictions, minimal innovation.
- Utah: Guardrails with room to innovate.
- Everywhere else: Some degree of “total freedom,” for now.
It won’t take long before we can compare outcomes. Which approach better serves public health, innovation, and individual choice?
Why Illinois’ Approach Misses the Point
Illinois lawmakers may think they’re protecting citizens from untested technology. But in doing so, they’ve closed the door on exactly the kinds of incremental, responsible deployments that could build public trust.
AI is already being used in mental health contexts—both formally, in pilot programs, and informally, by individuals who find comfort in speaking to a chatbot during a difficult moment. Some of these applications show genuine promise:
- Providing after-hours support when human therapists aren’t available.
- Helping underserved communities with limited access to mental health services.
- Offering multilingual, culturally adaptive conversation support.
The Illinois model blocks most of these possibilities before they can even be tested. I’ve found recent research from Dartmouth and Stanford to be particularly helpful. The Dartmouth study involved an experimental deployment of Therabot, a therapy chatbot, with volunteers diagnosed with mental illnesses. This study generally showed positive results in symptom relief. The Stanford study comprised two parts: the first investigated stigma across various AI systems, including Character.ai, and the second analyzed bot responses to severe symptoms like suicidal ideation. Collectively, these studies provide a comprehensive and balanced view of the initial risks and benefits associated with tailored therapy AI and companion AI. The results also suggest a qualitative difference between companion AI marketed as therapy and specialized therapy AI like Therabot.
A Better Way Forward
When it comes to AI and young people—or AI and mental health—the same principle should guide us: measure, don’t ban.
For teens, that means continuing to survey, track usage trends, and study potential harms before defaulting to restrictive policies that may stifle beneficial uses. For therapy, it means developing clear professional standards for AI integration—standards that preserve accountability without foreclosing innovation.
We’ve been here before with other technologies. Early fears about the telephone, television, and even video games often overstated the risks and underestimated the adaptability of both technology and society. That doesn’t mean the concerns were baseless—it means the policies that stand the test of time are the ones that balance risk mitigation with openness to new possibilities.
Why This Matters Now
AI development is moving faster than our ability to legislate thoughtfully. That creates a strong temptation for policymakers to “do something” quickly. But speed is not a substitute for wisdom.
The Common Sense survey shows us that, at least for now, fears about teens abandoning real human interaction for chatbots may be overstated. And the Illinois law shows us how quickly that fear can calcify into policy that does more harm than good.
If we want to get this right, we need:
- Better data — especially over time, so we can see trends rather than reacting to snapshots.
- Policy experiments — like Utah’s approach, which allows innovation within guardrails.
- A commitment to preserving beneficial uses — whether that’s a teen exploring a new technology or an AI tool expanding access to mental health care.
Bottom line: We should be watching both of these stories closely. How teens interact with chatbots will tell us a lot about AI’s role in everyday life—and how states regulate AI in sensitive domains will set the precedent for years to come. Let’s make sure both conversations stay grounded in data, not just fear.