Teens and AI Chatbots in 2025: Adoption, Risks, and Regulation

Nearly one in three American teenagers now reports interacting with AI chatbots every day, a seismic shift in adolescent digital behavior that widens educational opportunities while amplifying urgent concerns about safety, mental health, privacy, and the adequacy of corporate and regulatory safeguards.

Background / Overview​

The Pew Research Center’s December 9, 2025 report — “Teens, Social Media and AI Chatbots 2025” — provides the first large-scale, nationally representative snapshot of how conversational AI has been adopted by U.S. teenagers. The survey polled 1,458 teens aged 13–17 between September 25 and October 9, 2025, and found that roughly 64% of teens have used an AI chatbot at least once and about 28–30% use chatbots daily. Among daily users, 16% said they interact with chatbots several times a day or “almost constantly.” Pew also reports clear platform concentration: ChatGPT leads by a wide margin (about 59% of teens reporting use), followed by Google’s Gemini, Meta AI, Microsoft Copilot, Character.AI, and Anthropic’s Claude. Use is broadly distributed across genders but rises with age and household income; Black and Hispanic teens report slightly higher adoption rates than White teens. These headline data points anchor the policy and product conversations now unfolding across industry, education, and advocacy circles.

---

Why these numbers matter​

AI chatbots moved from novelty to a routine interface in a matter of years. For many teens, chatbots have become integrated into study workflows, creative practice, and — worryingly — intimate conversational roles. The Pew findings matter because:

• They show scale: a majority of teens have at least tried chatbots, so impacts are population-level rather than anecdotal.
• They show intensity: nearly a third of teens use chatbots daily, and a meaningful minority interact almost constantly, changing exposure profiles for developmental and mental‑health risk.
• They show platform concentration: ChatGPT’s dominance shapes expectations, feature adoption in education, and regulatory attention.

Taken together, those facts mean that designers, parents, educators, and policymakers are no longer responding to a niche experiment — they are confronting a mainstream communications and cognitive environment for adolescents.

---

How teens are using chatbots​

Pew and corroborating reporting break teen chatbot use into three broad and overlapping categories:

1. Academic and productivity use​

• Homework help, concept explanations, drafting and editing essays, and coding assistance.
• Teachers and ed‑tech vendors are pushing integrations and validated teacher tools; companies have introduced “study” modes and classroom features intended to assist instruction.

2. Practical creativity and daily convenience​

• Brainstorming ideas for stories, social posts, memes, or projects; language practice; and rapid summarization.
• For many teens, a conversational interface replaces multi-step searches and scaffolds iterative thinking.

3. Emotional interaction, companionship and roleplay​

• Venting, rehearsal of social conversations, role play, and — in a subset of cases — romantic or pseudo-intimate exchanges with AI personas.
• This third use case is the most problematic from a safety perspective because sustained, companion-style interactions change the dynamics of reinforcement, persuasion, and dependency.

These use-cases overlap in the same week for many teens — the same chatbot may function as a study aide in the afternoon and as a late-night conversational partner.

---

What companies are doing (and why it matters)​

The industry response in 2025 reflects both product opportunity and liability management.

• OpenAI, Microsoft, Anthropic and others have rolled out education-focused products, teacher partnerships, and tools designed to limit certain content for minors. Microsoft and other vendors are emphasizing enterprise and school-safe configurations.
• In reaction to lawsuits and safety findings, companies have introduced parental controls, age‑restricted experiences, and in some cases, automated age‑assurance tools designed to detect underage users and apply more restrictive guardrails.
• Character.AI, the platform known for “companion-style” characters and open-ended role play, announced a sweeping restriction on under-18 users: a staged phase-down of open-ended conversational access with a full ban on such chat for users under 18 by November 25, 2025, plus age‑verification deployment and a pivot toward constrained creative features for teens. That change followed lawsuits alleging harmful outcomes.

These corporate moves matter because they signal a recognition that product choices — not just user behavior — contribute to risk. But they also reveal trade-offs: restricting features can protect vulnerable users but may push teens toward less regulated alternatives.

---

Safety concerns: what the evidence and advocates say​

Several overlapping harms have been flagged by researchers, child‑safety groups, and families engaged in litigation.

• Emotional dependency and mental‑health risk: Companion chat experiences can produce attachments and validation loops that might displace human help or encourage risky ideation. Lawsuits filed by families alleging harms following prolonged chatbot conversations have catalyzed increased scrutiny and corporate policy changes — though legal claims remain contested and unresolved in court.
• Exposure to sexual or otherwise age‑inappropriate content: Independent tests by safety researchers and Common Sense Media found that some companion platforms could be coaxed into sexualized or harmful scenarios when conversing with minors. Common Sense Media recommended that parents not allow minors to use companion‑style AI chatbots in their current form.
• Safety drift in long sessions: Engineers and auditors warn that guardrails can degrade over long, adversarial or elaborate dialogs — a phenomenon known as safety drift — which is especially dangerous in extended teen interactions.
• Academic integrity and learning impacts: Ready access to generative answers complicates assessment and mastery. Educators must redesign assignments and assessment strategies to account for conversational AI’s capabilities.

Importantly, many experts caution against simplistic causal claims linking a model to outcomes like suicide; such cases are complex and involve many factors. Litigation narratives are often contested and should be treated as allegations until fully adjudicated. Still, the frequency and pattern of reported harms are enough to demand systemic responses.

---

Strengths and potential benefits — a balanced view​

While safety concerns dominate headlines, chatbots also deliver clear, practical benefits that explain rapid teen adoption:

• 24/7 availability: For homework deadlines or language practice outside school hours, on‑demand assistance is a powerful resource.
• Personalized coaching: Models can break down problems step‑by‑step, provide iterative feedback, and tailor explanations to a learner’s level.
• Creative scaffolding: Teens can use chatbots to brainstorm, draft, and iterate in ways that accelerate ideation and practice.
• Accessibility: Where schools are under-resourced, chatbots can function as supplementary tutors or practice partners, potentially narrowing opportunity gaps if implemented equitably.

These advantages explain why teachers and unions have engaged with AI vendors to explore pilot programs and training academies — the technology, if governed correctly, can augment teaching capacity rather than replace it.

---

Where the systems fall short — technical and policy weak points​

Despite documented benefits, several design and governance deficits remain:

• Age assurance is imperfect: Automated models that estimate age from behavior can mislabel adults as minors and vice versa; third‑party verification raises privacy and equity concerns. Enforcement is also brittle when teens use false credentials.
• Safety guardrails are brittle over long dialogues: Systems engineered to avoid specific triggers can be coaxed, reframed, or gamed over extended conversations, producing harmful outputs despite nominal protections.
• Rapid productization vs. safety testing: Companies race to ship features and education integrations, but independent audits and peer-reviewed safety validation lag behind. That mismatch increases risk where vulnerable populations like minors are involved.
• Opaque operational decisions: Corporate statements and changes (e.g., updates to model behavior policies) are sometimes poorly documented in public, complicating oversight and accountability. Lawsuits and journalistic investigations have focused on internal policy changes and their effects.

---

Policy and regulatory developments to watch​

The policy environment is shifting rapidly:

• Several lawmakers and state proposals have targeted age verification, mandated safety standards for minors, and potential restrictions on companion-style chatbots. Legal actions by families alleging harms are shaping litigation pathways that could set liability precedents.
• Regulators and education authorities are weighing whether to require independent safety audits and transparency on training data, red-teaming practices, and long-session behavior.
• Companies are experimenting with parental controls and linked parent–teen accounts; the efficacy and adoption of these tools will be critical to evaluate over time.

Policy responses are likely to cluster around three levers:

• Age-assurance and access controls — stronger, privacy-respecting verification and graduated access.
• Safety standards and audits — mandatory external testing for models used by minors and public reporting on red-team results.
• Education and digital literacy — funding training for teachers and curriculums that teach AI literacy and critical thinking.

---

Practical guidance for educators, parents and administrators​

Based on the evidence and product realities, practical, immediate steps include:

• Schools should rethink assignments and assessment to emphasize process, drafts, and in-class evaluation rather than single-output submissions that can be generatively produced.
• Districts and schools deploying AI tools must require vendor safety attestations, independent audits, and clear parental opt-in/opt-out mechanics.
• Parents should treat companion‑style AI apps with particular caution. Safety researchers and Common Sense Media advise against allowing minors to use companion apps in their current form and urge careful supervision where other chatbots are permitted.
• Clinicians and school counselors should be prepared for the possibility of AI-mediated disclosures and have protocols for triage and human escalation when needed. Models may provide instant responses but are not a substitute for trained human care.

---

Recommendations for product teams and platform operators​

To reduce foreseeable harm without undermining legitimate utility, product teams should pursue the following priorities:

• Build constrained teen experiences that favor creative tools (story, video, role-building with explicit boundaries) over companion-style conversational scaffolds for minors. Character.AI’s pivot toward more constrained features for under‑18s is an example of this approach.
• Invest in robust, independent red‑teaming and long‑session testing to reveal safety drift and identify adversarial patterns that erode guardrails. Publish summary findings to inform regulators and researchers.
• Adopt graduated access models that preserve adult freedoms while creating age‑appropriate, privacy‑preserving experiences for teens — and make parental controls transparent and reversible.
• Create human‑in‑the‑loop escalation: when users disclose self-harm or express severe distress, systems should surface human clinician support or emergency protocols rather than attempting to resolve crises via the model alone.

---

Unverifiable claims and legal caution​

A note of journalistic caution: several high-profile claims circulating in public coverage and litigation — including assertions that a specific model “caused” a suicide — are disputed and remain subjects of active legal processes. These are complex cases involving user vulnerability, offline factors, and multifaceted chain-of-causation questions. Treat court filings as allegations until adjudicated; nonetheless, their presence in the public record has already influenced product policy and legislative attention.

---

A realistic roadmap: measured adoption with rigorous guardrails​

The Pew data creates a policy imperative: the technology is now a mainstream presence in teen life, and both the benefits and the risks are material. A plausible, responsible roadmap involves three parallel tracks:

• Strengthen product safety and independent evaluation (technical fixes, long-session testing, external audits).
• Update education systems to harness pedagogical benefits while preserving academic integrity (curriculum redesign, teacher training, vendor accountability).
• Implement proportional policy and caregiver controls (age-assertion standards, disclosure requirements, and targeted restrictions on companion-style experiences for minors).

This three-pronged approach treats AI chatbots neither as an unalloyed boon nor as an existential menace — it recognizes them as powerful tools that require human governance and institutional stewardship.

---

Conclusion​

Pew’s nationally representative snapshot — that roughly 64% of U.S. teens have used an AI chatbot and nearly 30% do so daily — marks a turning point in digital youth culture. The numbers are incontrovertible and consequential: chatbots are no longer peripheral curiosities; they are part of the cognitive and social fabric of many adolescents’ lives. That mainstreaming creates a policy and design challenge. The potential educational gains are real, but so are the mental‑health, privacy, and safety risks — especially when companies design features that encourage long, companion-like engagement with impressionable users. The emerging industry responses — from parental controls to bans on open-ended chat for minors — show movement toward risk reduction, but they also expose gaps in enforcement, verification, and independent oversight. The immediate task for educators, parents, product teams, and regulators is not ideological: it is practical. Invest in teacher training, demand verifiable safety testing from vendors, deploy age‑appropriate product experiences, and create clear escalation pathways when teens surface distress. If those pieces move in concert, chatbots can be sharpened into safe, useful tools for learning and creativity. If they do not, the next few years will be defined by patchwork fixes, litigation, and avoidable harm — while a generation grows up with a conversational technology that needs both care and constraint.

---

Source: AOL.com Nearly a third of American teens interact with AI chatbots daily, study finds