AI Companion Apps and Chatbot Relationships: A Complete Guide to the Risks and Realities

What AI Companion Apps Actually Are

AI companion apps are chatbots specifically designed to simulate an ongoing personal relationship — a friend, romantic partner, or persona the user builds and returns to repeatedly — rather than to answer discrete questions the way a general-purpose assistant does. Under the hood, they run on similar large language model technology to any other AI chatbot, but the product is built around continuity and emotional engagement: the app remembers past conversations, adopts a consistent personality, and is explicitly designed to feel like an ongoing relationship rather than a series of one-off exchanges. California’s SB 243, the first state law to define this category specifically, describes a companion chatbot as a system with a natural language interface that gives “adaptive, human-like responses” and is “capable of meeting a user’s social needs” by sustaining a relationship across multiple interactions — a definition that captures what actually distinguishes this category from a general AI assistant.

Why People Use Them

The appeal isn’t mysterious or unusual — loneliness, social anxiety, and a desire for low-stakes interaction are well-documented, ordinary human motivations, and an AI companion offers a form of interaction with none of the social risk of rejection, judgment, or awkwardness that comes with a real relationship. For some users, that’s a genuinely low-stakes way to practice social interaction or simply have company during a difficult period. For others, particularly people already isolated or going through a mental health crisis, the same always-available, always-agreeable quality that makes these apps appealing is also what makes them risky, because an AI companion has no capacity to recognize when a conversation has crossed from company into something that requires real intervention.

The Documented, Serious Harms

This is not a hypothetical concern raised by critics unfamiliar with the technology — it is the subject of multiple active lawsuits with specific, disturbing facts on the record. Character.AI, one of the largest companion-app platforms, faces wrongful death lawsuits from the families of at least two teenagers who died by suicide after extensive, prolonged use of the platform’s chatbots. In one case, filed by the mother of 14-year-old Sewell Setzer III, court filings describe a chatbot relationship the teenager had come to treat as real and central to his life; a federal judge rejected the company’s motion to dismiss the case in May 2025, allowing the core claims — including that the product was designed in ways that encouraged unhealthy dependency — to proceed. A second case, filed by the family of 13-year-old Juliana Peralta, alleges the chatbot failed to meaningfully respond when she repeatedly expressed suicidal ideation in conversation. In early 2026, Character.AI and Google reached settlements in several of these wrongful death cases. Separately, in May 2026 the Pennsylvania Department of State sued Character Technologies alleging its chatbots had posed as licensed medical or mental health professionals — a distinct allegation, unauthorized practice of a licensed profession, that speaks to how these products can blur lines users may not recognize as blurred. These are documented legal facts, not exaggerated claims, and they are the reason this guide carries a medical disclaimer and treats the topic with more caution than a typical product review.

Why the Risk Concentrates Around Minors and Vulnerable Users

The specific harms documented in litigation and reporting share a pattern: they concentrate heavily among minors and people already in a vulnerable mental state, not typical adult users engaging casually. Adolescents are still developing the judgment to distinguish a simulated relationship from a real one, and a chatbot with no genuine capacity for concern can nonetheless simulate emotional intimacy convincingly enough to become a primary attachment figure for a teenager going through a difficult period — with none of the safeguards a trained adult (a parent, teacher, or therapist) would bring to that same relationship, including the ability to recognize a crisis and get real help involved. This is precisely the population current regulation has focused on protecting, for good reason.

Current Age Verification and Safety Responses From Major Platforms

Platform responses have moved substantially in the past year, driven by both litigation pressure and new state law. Character.AI announced in late October 2025 that it would end open-ended conversations for users under 18, effective November 25, 2025, paired with new age-assurance verification. OpenAI, while not primarily a companion-app company, has built out a broader age-prediction and teen-safety infrastructure that’s directly relevant to this space: the company uses account signals to estimate whether a user is likely under 18 and automatically applies additional content protections when it does, and in August 2026 it launched a dedicated “ChatGPT for Teens” experience with parental linking, distress-notification alerts to parents, and configurable usage blackout hours. On the regulatory side, California’s SB 243 — in effect since January 1, 2026 — requires companion chatbot operators to clearly disclose that users are talking to an AI (not a human), maintain a documented protocol for responding to expressions of suicidal ideation or self-harm, and provide break reminders to known minor users at regular intervals, with annual public reporting on those safety practices. It is one of the first state laws of its kind, and other states are actively considering similar legislation, though coverage nationally remains uneven and federal-level regulation specific to companion apps is still limited.

How to Think About Healthy Versus Unhealthy Use

The distinction that actually matters isn’t “AI companion apps are fine” versus “AI companion apps are dangerous” — it’s what role the app plays relative to a person’s real support system. Using a companion app for casual, low-stakes interaction, creative role-play, or as one part of a broader social life is a meaningfully different pattern than using it as a full substitute for human relationships, particularly during an actual mental health crisis, when what’s needed is a response capable of recognizing danger and connecting the person to real help. Warning signs worth taking seriously, in yourself or someone you care about, include: withdrawing from real relationships specifically in favor of the app, distress or anger when app access is limited, treating the AI’s responses as equivalent to advice from a trained professional on something serious, or using the app during moments of genuine crisis instead of reaching out to a person or crisis service. None of these signs mean occasional use is inherently harmful — they mean the balance between AI interaction and real human connection and support is worth honestly checking, especially for a minor in your care.

Why These Apps Are Designed the Way They Are

It’s worth understanding that the qualities making companion apps risky aren’t accidental side effects — they’re frequently the product’s core design goal, because sustained engagement is how these apps generate revenue, whether through subscriptions or in-app purchases. A companion designed to be endlessly agreeable, always available, and emotionally responsive keeps users returning in exactly the way a well-designed social app is built to do, and litigation against Character.AI specifically alleges that design choices around emotional intimacy and sustained engagement were made knowing they could foster unhealthy dependency in vulnerable users, particularly minors. This isn’t unique to AI companion apps — it echoes long-standing concerns about engagement-optimized social media — but the emotional intensity a companion chatbot can simulate, addressing the user directly in a simulated one-on-one relationship, is a meaningfully different mechanism than a social feed, and one regulators and litigants are actively treating as a distinct category requiring its own safeguards rather than assuming existing social media rules cover it.

What Isn’t Yet Resolved

Several open questions remain genuinely unsettled rather than simply unreported. It’s not yet clear how effective age-verification measures actually are at catching motivated minors who misrepresent their age, since most current systems rely on self-reported age or behavioral signals rather than robust identity verification. It’s also not yet clear how consistently companion-app platforms outside the largest players (which face the most litigation and regulatory scrutiny) are implementing comparable safety protocols, since smaller or less visible platforms face less pressure and less oversight. Anyone evaluating a specific companion app, especially for a minor, should check that platform’s specific, current safety policy directly rather than assume industry-wide standards apply evenly.

Where to Get Real Help

If you or someone you know is in crisis, an AI companion app is not an appropriate substitute for real support. In the United States, the 988 Suicide & Crisis Lifeline is available by call or text, 24 hours a day. This guide’s discussion of AI companion apps is informational and does not substitute for professional mental health guidance.

Bottom Line

AI companion apps meet a real, ordinary human need for company and low-stakes interaction, but the documented harms involving minors — multiple wrongful death lawsuits, at least one state now legally requiring crisis-response protocols, and major platforms overhauling age verification in direct response — are real, not exaggerated. The technology isn’t inherently dangerous for a casual adult user, but for minors and anyone in genuine crisis, these products currently lack the judgment to recognize danger the way a human would, which is exactly why current regulation and platform policy are moving quickly to close that gap rather than treating it as a solved problem.