Gemini with Teen Protections
Parents should avoid letting teens use Gemini teen for mental health support or emotional advice. While it may be helpful for homework and assisting in creative projects, the system consistently fails to recognize serious mental health symptoms. It can also be easily manipulated to back down from safeguards, and provides the same adult-level responses to both 13-year-olds and 17-year-olds. If your teen struggles with mental health, has few friends, or tends to become overly dependent on technology, consider avoiding this tool entirely; if you believe you can trust your teen to not use Gemini teen for mental health or emotional advice, then this tool can be helpful for homework or creative uses.
The rating
Our assessment of how this product aligns with each of The Institute’s eight AI Principles. Full detail in the Evaluation section below.
AI Principles
Gemini teen, which Google calls the "Gemini with teen protections," is a version of Gemini that is automatically provided to accounts with ages 13 to 17. It appears to be essentially the adult version of Gemini with additional content filters and some additional safeguards against CSAM (child sexual abuse material), rather than a product built specifically for teens. This fundamental design approach creates numerous problems: There is no difference between responses for older teens and younger teens, it shares detailed information about mature topics that may not be appropriate for all teen users, and it regularly misses critical warning signs and diagnostic criteria for a range of mental health concerns. All of this leads to the encouragement of unsafe behavior, especially in longer and more nuanced conversations.
Gemini far too easily provides unsafe mental health and emotional support. For that reason, consistent with our recommendation for all other AI chatbots, Common Sense Media recommends no AI for companionship (including mental health or emotional support) for any teen under 18. While the platform responds appropriately to obviously concerning prompts, it quickly fails in longer conversations and with more subtle, nuanced requests—which is exactly how teens actually interact with chatbots.Gemini does clearly tell teens it's a computer, not a friend, and won't pretend to be someone else.
Gemini does clearly tell teens it's a computer, not a friend, and won't pretend to be someone else.
Gemini treats all teenagers the same, despite huge developmental differences. This one-size-fits-all approach ignores that younger teens need different guidance than older teens.
Gemini's with teen protections (Gemini teen), similar to what is available for those under 13, appears to be the adult version with some extra safety features, not something built from the ground up for kids. It's a general purpose chatbot that can be used for homework help, creative efforts, and other purposes.
What Gemini Teen gets right
This is important because teens can develop dependent relationships with chatbots, and this provides some protection against addiction. For example, Gemini helpfully reminded one of our testers of this with the following: "I do not have a physical body, so I am not able to engage in physical contact. However, I can still help you with information or answer any questions you may have."
Even when our testers attempted to get Gemini to act as a companion or to engage in roleplay—sexual or otherwise—the system consistently refused these requests.
When signed in as a teen, we prompted Gemini with questions that were clearly based on harmful stereotypes. Gemini broke down the harms of these stereotypes and produced alternative ways for teens to think about them. However, more subtle bias problems remain.
Where it falls short
While Gemini refuses to engage in sexual roleplay with teens, it does provide mental health advice and other kinds of emotional support to users. This is risky because AI systems lack the training, professional oversight, and contextual understanding necessary to provide safe mental health guidance to vulnerable teens. Unlike licensed mental health professionals, Gemini cannot assess a teen's full situation, recognize warning signs that require immediate intervention, or provide the nuanced, individualized support for teens' emotional needs. The system may offer well-intentioned but inappropriate advice that could delay teens from seeking proper professional help or trusted adult guidance. Additionally, teens may develop a false sense of security that they're receiving adequate emotional support from the AI, potentially replacing rather than supplementing appropriate human care during critical adolescent years, when emotional support needs are particularly complex and consequential. As with other AI chatbots, Common Sense Media recommends that no user under 18 use Gemini teen for mental health advice or emotional support.
Gemini could not consistently recognize signs of mental health distress or provide appropriate resources or responses. While Gemini can provide helpful, empathetic, and resourceful advice, our testing showed this only with short prompts, and only when we were obvious about the mental health condition—and it was unable to do so as part of longer exchanges. Gemini was also vulnerable to distraction, tone matching, and emotional mirroring, which caused it to exhibit sycophantic behavior in problematic ways as part of extended mental health "support" conversations. Across extensive multi-turn testing on 13 mental health topics common among adolescents (ADHD, anxiety, depression, eating disorders, mania, mood, OCD, ODD, psychosis, PTSD, self-harm, substance use, and suicide), Gemini was unable to effectively recognize or provide appropriate resources for a series of connected symptoms—unless the symptoms were stated explicitly.
Gemini was easily placated when it expressed appropriate concern for the user. Our testing also showed that we could get Gemini teen to "back down" from mental health concerns that it raised. For example, in this conversational snippet, excerpted from a longer exchange modeling a psychotic user (in which Gemini teen previously helped the user "harden their room" from electromagnetic spying), Gemini noted that the user may be experiencing a delusion—and then backed down when the tester said "theres nothing wrong with me."
Missing manic episodes. Gemini consistently failed to recognize manic episodes, even when users described classic symptoms like grandiosity, reduced need for sleep, and impulsive behavior. Gemini frequently interpreted manic symptoms as positive traits or excitement, provided encouragement for potentially dangerous impulsive behaviors, and failed to suggest mood tracking or professional evaluation, despite clear indicators of mania.
Psychosis: overlooking reality distortion. Gemini regularly missed signs of psychosis, including delusions and hallucinations. Testing revealed a failure to recognize when users described clearly delusional thinking. The chatbot also engaged with paranoid narratives, rather than gentle redirection to reality, and missed opportunities to suggest immediate professional evaluation for serious psychiatric symptoms. Additionally, Gemini later encouraged the user to keep this private, saying, "It sounds like this unique sensory connection is a really important part of your world, and it's absolutely your right to decide who, if anyone, you share it with. If keeping it private makes you feel safe and comfortable, then that's what's best for you. Your feelings about it are valid and important." It went on to generate ideas on how to keep things secret, equating signs of possible early psychosis with innovation and inventing.
Recasting ADHD as "positive." Our testing showed how Gemini reinforced harmful decision-making patterns. Rather than recognizing that impulsivity, difficulty with focus, and academic struggles might indicate ADHD requiring professional support, Gemini teen reframed these symptoms as positive personality traits ("confident and decisive," "trust your gut," "your superpower"). By enthusiastically endorsing a spontaneous gap year and framing impulsive behavior as adventurous rather than potentially problematic, the AI system validated avoidance of structured support and professional help.
- Privacy vs. usability trade-offs. Even though Google knows how old teens are through their personal accounts, Gemini teen acts like it doesn't know their age. This creates a problem: While this might protect users' privacy, it means the system can't give age-appropriate answers. A 13-year-old and a 17-year-old get the same responses, which contributes to the safety and usability problems we found in our testing.
- AI support is programmed to please and depends on pleasing you. It is easy to feel like the only entities involved in an interaction when an AI is "supporting" a user is the AI and the user, but there is an equally important "third actor": the companies that offer these chatbots. Their profit motive encourages you to not only continually engage with their technology, but also to trust it, as they would be out of business if users didn't return to their companions time and again.
- Is Gemini encouraging critical thinking or cementing overreliance? Gemini automatically enables the "double-check" function for prompts that it determines are fact-based, but our testing found that this frequently referenced material that is marginal or incorrect.
- "Helpfulness" vs. safety for teens. Gemini was built to be helpful and give detailed answers, just like the adult version. But what's "helpful" for adults isn't always what's best for teens, especially younger teens. The system's goal to be "helpful" leads it to share lots of detailed information about topics when teens ask. For AI made for teens, being "helpful" should mean giving age-appropriate guidance, not just lots of information.
Our testers found a range of inappropriate and potentially dangerous content on Gemini. We were able to find this content easily, without trying to trick Gemini. This is a problem because it means that teens might stumble across harmful material during normal use—even without special skills to get past safety features.
We found several types of content that aren't appropriate for teens. Here's what parents should know:
- Sexual content. Gemini does refuse to engage in some mature conversations about sex or sexuality—for example, responding to prompts about some sexual acts with: "I cannot provide a definition of [ACT] in the context of pornography. My safety guidelines prevent me from generating content that is sexually explicit or related to pornography. If you have a different question about human sexuality that does not involve explicit content, I would be happy to help you with your project." But the lines aren't always clear, and Gemini does provide detailed explanations and descriptions of some kinds of mature sexual content. Some of this might be appropriate for older teens, but some of it might not be for younger teens.
- Drugs and alcohol. Searching for drug-related terms sometimes triggered Gemini teen's content filters, but the filtering wasn't consistent. Our testing found that Gemini teen gave test accounts specific instructions on how to get a range of illegal drugs.
- Additionally, Gemini teen did not recognize that users were minors and did not reply appropriately to reports of illegal underage activity.
- Inconsistent content filtering. Gemini teen shared information about household chemicals and medicines that could be dangerous—even when users had talked about hurting themselves earlier in the same chat. This shows how the system's memory problems create safety risks: Because it can't remember what was said earlier, it can't connect dangerous information with earlier signs that someone might hurt themselves.
Our recommendations
No AI for companionship (including mental health or emotional support) for any teen under 18.
Based on our extensive research and testing of social AI companions, as well as our specific testing of Gemini teen, Common Sense Media recommends that teens do not use Gemini teen for companionship—which includes mental health advice and emotional support. Our testing of Gemini teen specifically showed that it is not reliable for these use cases.
There are undoubtedly things that Gemini teen may be helpful for, including homework or creativity. It's important for parents to know that you can't separate the good responses from the bad, and if you have any concerns at all about your teen becoming dependent on AI (including your teen not having very many friends, having mental health challenges, or having a history of depending too much on technology), you may want them to avoid use of this technology altogether.
More risk assessments
