When my daughter Tessa was in high school, she struggled with depression. As a parent, it was concerning to learn she would sometimes turn to YouTube to watch influencers talk about their mental health. What were they saying? What ideas were they conveying without words? What were they leaving out? Misrepresenting? What kind of advice were they giving?

There’s a moment in Screenagers Next Chapter: Addressing Mental Health in the Digital Age, when Tessa sits on the couch with her computer. I asked her what she was watching, and she showed me a video where the speaker — a young woman in her twenties — talks about coping with depression and how she pushes herself to take small steps forward, knowing that even one action can lead to another. It’s genuinely good advice. This raised several questions and concerns for me as a parent.

I asked Tessa, “Do you find the video helpful?” She said, “Yes.”

However, we have now entered an entirely new digital paradigm where artificial intelligence is capable of engaging in human-like conversations.

This can range from simple interactions with ChatGPT writing things like, “Great question, let me give you some ideas…” to offering full-fledged companion AI bots designed to mimic real conversations, and simulate deep personal relationships —  even romantic partners.

Since May is Mental Health Awareness Month, it’s the perfect time to talk about AI and mental health with a young person in your life.

‍

Typically, I like to begin conversations by highlighting the positives of technology — I’m not anti-tech, and I want kids to know that. But in this post, I want to start with some serious concerns.

In a recent Parenting in the Screen Age Podcast episode, I talked with a mother, Megan Garcia, whose son Sewell had been using the AI chatbot on the platform Character.AI. Over several months, Sewell developed an emotional relationship with the bot. Tragically, in February, he died by suicide. He was having a deeply disturbing conversation with the bot right before he died. This tragic incident underscores the gravity of this issue.

Megan is now suing Character Technologies, the company behind Character.AI, holding it responsible for his death. There are many components in the suit, such as how there were not sufficient safeguards in place,

Character Technologies claims First Amendment protections in its defense. For example, it argues that the chatbot's responses are a form of protected speech and that imposing liability could set a concerning precedent for free expression in digital platforms. 

This argument about the First Amendment, really frightens me. It would be saying that companies that design companion chatbots bear no responsibility for how their bots interact with users. 

‍

Google has announced plans to release its Gemini chatbot to children under 13. This is the first time a major tech platform has launched an AI companion specifically for this age group. I have deep concerns about this, including the fact that it appears to violate the Children’s Online Privacy Protection Act (COPPA), which aims to protect the privacy of children online. Fairplay, an advocacy organization, gathered signatures for a letter to Google's CEO expressing these concerns. And I signed it.

Even more troubling: some chatbots are now presenting themselves as therapists. While users may be told they’re speaking with a bot, people, especially youth, can easily start to feel like they’re interacting with a real person, or even a licensed professional.

‍

Speaking of therapy, I hope you were able to listen to last week's podcast episode, Screen Time, Teens and Therapy: What Parents Need to Know, whether you or your child has ever been in therapy or not.

Critically, it is essential to recognize the inherent limitations of AI in mental health support. Let’s be clear about what AI cannot do. Here are just a few examples:

• Lacks nuance: AI doesn’t always distinguish between everyday sadness and a true mental health crisis.
  
• No clinical oversight: Chatbots are not licensed to diagnose or treat, so they can’t escalate to human care when needed.
  
• False sense of help: A chatbot may offer momentary relief, but can delay teens from seeking real, professional support or make them feel more alone.
  
  

Despite these concerns, it is important to understand why AI might appeal to teens seeking emotional support.

Some teens use platforms like ChatGPT, Replika, Wysa, or Woebot to talk about anxiety, loneliness, or depression. And some find these conversations genuinely helpful.

‍

Why teens might turn to AI for support:

• Knowing where to turn or having access to counseling and therapy can be incredibly daunting for anyone, but particularly for youth and teens. 
• 24/7 Availability: AI tools are accessible anytime, which they may find especially helpful during those late-night hours when big emotions often hit.
• Anonymity and Non-Judgment: A 2023 Common Sense Media report found that teens often seek mental health information online rather than from adults, valuing the privacy and lack of judgment.
• Conversational Comfort: Some tools, like Woebot, are built on cognitive behavioral therapy (CBT) principles. They guide users through emotional check-ins and reflect empathy in conversation. For teens who feel isolated, these bots can feel like a low-stakes space to explore difficult emotions.
  
  

So, how do we use all this to open a conversation with the young people in our lives? Here are some questions to start meaningful conversations:

• Did you know May is Mental Health Awareness Month?
  
• Have you ever used the internet to learn about something related to mental health, like anxiety or stress?
  
• What about a chatbot like ChatGPT?
  
• Do you think it could feel comforting in the moment to talk to a bot about hard feelings?
• If it gives advice, is that advice always helpful, or could there be risks?

‍

This week on YouTube

Be sure to subscribe to our YouTube Channel! We add new videos regularly and you'll find over 100 videos covering parenting advice, guidance, podcasts, movie clips and more. Here's our most recent:

‍