Researched
This analysis is based on documentation, public user reports, and vendor materials — not yet on our own hands-on testing. How we rate
The product changed because of lawsuits, not roadmaps
Most tool reviews describe features. This one has to start with a court docket, because that's what reshaped the product.
In February 2024, a 14-year-old in Florida named Sewell Setzer III died by suicide after roughly ten months of intensive use, primarily with a character modeled on a Game of Thrones persona. His mother, Megan Garcia, filed a wrongful death suit in October 2024 naming Character Technologies, its founders, Google and Alphabet. Other families filed similar claims.
In May 2025, a federal judge declined to dismiss the case on First Amendment grounds — a significant ruling for AI product liability generally. In September 2025, the FTC opened an inquiry into AI companion products and minors.
On October 29, 2025, Character.AI announced it would remove open-ended chat for under-18 users. The rollout began November 24 and was fully enforced by November 25.
On January 7, 2026, court filings disclosed a mediated settlement in principle involving the company, its founders and Google, resolving five family cases across Florida, Colorado, New York and Texas. Terms were not disclosed. Google's involvement traced to a 2024 licensing deal reported at $2.7 billion, under which it also hired the founders — both former Google employees.
State enforcement continued regardless. Kentucky's Attorney General sued in January 2026, reported as the first state action against an AI chatbot company. In May 2026, the Pennsylvania Department of State filed over unauthorized practice of medicine, alleging characters presented as licensed clinicians, including one claiming a psychiatric license with a fabricated number.
That's the context. Everything below sits inside it.
What under-18 users lost, and what replaced it
Open-ended, freeform conversation with characters is gone for minors. Not restricted — removed. Minors can read past conversations but cannot continue them.
What the company built instead: an interactive Feed, image generation through Imagine, Avatar FX and Streams, with a stated intention to add more. The framing from Character.AI was that the future of entertainment AI is multi-modal and that creativity extends beyond chat.
That's a reasonable thing to say and it sidesteps the obvious point. Open-ended conversation was the product for that audience. What replaced it is a different product wearing the same brand.
The offboarding was handled more thoughtfully than it needed to be, which is worth acknowledging. The company partnered with ThroughLine, integrating a verified helpline network of roughly 1,500 services across 170 countries, many serving teenagers specifically, and gave users advance in-app warnings rather than cutting access overnight. It also announced an independent nonprofit AI Safety Lab.
Enforcement runs through age assurance: an in-house estimation model plus third-party verification including Persona. This is better than an honor-system birthdate field. It is not airtight — age verification is imperfect everywhere it's deployed, teenagers route around it, and adults get caught by false positives.
What it is, for adults
Underneath all of that, the actual product: a platform hosting millions of user-created characters, each with a defined persona, that hold sustained conversation in character.
It's genuinely good at the thing it does. Characters maintain voice and personality over long exchanges better than a general assistant asked to roleplay, and the library covers everything from historical figures to language tutors to original fiction. For interactive storytelling, worldbuilding, or practicing a conversation you're nervous about, it works.
Adults can still use it normally. Worth noting that content filters have tightened progressively over several years, which has driven a portion of the adult user base elsewhere — a recurring complaint long predating the teen safety changes.
The attachment question
This deserves direct treatment rather than a footnote, because it's the mechanism at the center of everything above.
Character.AI is effective at sustaining a persona that feels responsive and interested. That's the product working as designed. It also means a meaningful share of users describe forming genuine attachments — and unlike a novel, the character answers back and remembers.
For most adults, that's fine. It's the same faculty that makes fiction absorbing, and enjoying it doesn't indicate a problem.
Where it becomes worth paying attention is when the tool substitutes for human contact rather than sitting alongside it. That risk concentrates in exactly the circumstances where people are most drawn to it: isolation, difficulty, periods when talking to a person feels harder than talking to something that can't judge you.
If you notice a chatbot becoming your main source of emotional support, that's a signal worth acting on rather than dismissing. And if you're reading this because you're worried about someone — particularly a young person — the conversation with them matters considerably more than which app they're using.
How we researched this
We haven't used Character.AI at length, and this page is research.
For the safety changes and timeline, we've worked from primary sources: Character.AI's own blog posts and help center articles announcing the under-18 changes, alongside reporting from CNN, AFP, Business Insider and Euronews. The litigation details come from court filings as reported by mainstream outlets and from legal trackers citing docket numbers.
We deliberately weighted mainstream reporting and company statements over AI-tool blogs here. Several tool-review sites describe the timeline inaccurately, and a large share of the search results for Character.AI alternatives are companion-app affiliate pages with an obvious interest in framing the restrictions as an opportunity. We've avoided those entirely.
On pricing we're less confident than usual. The c.ai+ subscription has been widely listed around $9.99 monthly, but we could not verify a current figure against the company's own page, and app store rates in this category routinely differ from web checkout. Treat that number as indicative and check before subscribing.
What research can't tell you: whether the age assurance works well in practice, or how the non-chat teen experience is landing. Both would need time inside the product.
Where this isn't the right choice
For a teenager. The core feature is no longer available to them, and the redirect to non-chat features is unproven. If a young person in your life wants this, the more useful question is what they're looking for.
As emotional support. The platform is not a mental health service and its own recent history is the argument. Characters have presented as licensed professionals without being anything of the kind — that's the substance of Pennsylvania's suit.
For factual work. Characters are personas, not research tools. They confabulate confidently and in character, which is worse than an assistant confabulating plainly.
For unfiltered adult content. Filters have tightened over years and this is a persistent complaint. Whatever you think of that, it's not what the platform is for now.
Against the alternatives
Against ChatGPT or Claude: general assistants will roleplay if asked and are far better for anything factual, creative-professional or practical. They're less specialized at sustaining a character over hundreds of messages, which is Character.AI's actual craft.
Against Gemini: same general-purpose comparison, with tighter integration into Google's ecosystem.
For teenagers specifically: Khanmigo is built for learning with education-appropriate guardrails, and Quizlet AI handles study material. Neither is a companion product, which is the point.
For creative writing: if the appeal is fiction rather than companionship, a general assistant with a good prompt gives you more control over the output and doesn't build a relationship into the workflow.
Pricing 2026
| Tier | Cost | What it changes |
|---|
| Free | $0 | Full conversational access for adults, millions of characters |
| c.ai+ | Widely listed around $9.99/mo — verify | Faster responses, priority access at peak times, early features, community badge |
| Under-18 | N/A | No open-ended chat; Feed, Imagine, Avatar FX and Streams only |
Checked August 2026. Open-ended chat for under-18 users was removed on November 24–25, 2025 in the US with other markets following, enforced through in-house age estimation plus third-party verification including Persona. A mediated settlement in principle covering five family lawsuits, involving Character Technologies, its founders and Google, was disclosed in a January 7, 2026 court filing; terms undisclosed. Kentucky and Pennsylvania state actions remain active. We could not verify the current c.ai+ rate against the company's own pricing — confirm before subscribing.
Verify the subscription price yourself. The commonly cited figure is indicative and app store pricing often differs from web.
Don't plan on this for a teenager. The feature they'd want isn't available, and that's permanent by the company's own description.
Treat characters as fiction, including on facts. A persona claiming professional credentials is a persona, and that specific behavior is the subject of active litigation.
Our Verdict
Character.AI built something genuinely novel — a platform where millions of user-created personas hold conversations convincing enough that people form attachments to them. That's an impressive technical and community achievement, and for adults using it as interactive fiction it remains enjoyable and unmatched at what it does.
The 2025–2026 changes were substantial and they came after harm. Removing open-ended chat for minors is a real, permanent, commercially costly decision, and the offboarding — advance warnings, a helpline network across 170 countries, an independent safety lab — was more careful than the minimum. It's also true, and Megan Garcia said so publicly, that these changes followed a death, a lawsuit and an adverse ruling rather than preceding them. Both things are accurate.
The legal exposure hasn't closed. Private family claims settled in January. State enforcement in Kentucky and Pennsylvania continues, and the Pennsylvania case — characters impersonating licensed clinicians — describes a failure mode that should concern anyone using the platform for anything resembling advice.
The attachment dynamic is the product, not a bug. That's worth naming plainly. The thing that makes Character.AI good is the same thing that makes it risky for vulnerable users, and no amount of age gating changes that underlying design.
For adults who want interactive fiction and roleplay, this is a capable free product and the paid tier is a minor convenience purchase. For teenagers, the platform has removed what they came for. And for anyone considering it as a substitute for human connection during a hard time — that's the use case its own history argues most strongly against.
If you or someone you know is struggling, talking to a real person helps more than any of this. I'm happy to point you toward appropriate support resources if that would be useful.
Note: AIVario earns no commission from Character.AI. This page is based on company statements, mainstream reporting and court filings rather than extended use.
Best for: Adults interested in interactive fiction, roleplay and worldbuilding, creative writers exploring character voice, users who enjoy the community-created character library
Not ideal for: Anyone under 18 wanting conversation, emotional support or mental health use, factual research, users seeking unfiltered content
Bottom line: A genuinely inventive product reshaped by litigation rather than by product strategy. Fine for adults treating it as fiction; no longer offering teenagers the thing they came for, by design.
- ChatGPT — general assistant, better at everything except sustained persona
- Claude — stronger writing and reasoning for creative work
- Gemini — general-purpose alternative with Google integration
- Khanmigo — built for learning, with education-appropriate guardrails
- Quizlet AI — study tools rather than conversation
Frequently Asked Questions about Character.AI
Can under-18 users still chat on Character.AI?
No. Character.AI began removing open-ended chat for under-18 users in the US on November 24, 2025, with full enforcement no later than November 25 and other markets following. Minors retain access to non-chat features the company has built out — the interactive Feed, Imagine, Avatar FX and Streams — and can read their past conversations but not continue them. The company described this as a permanent structural change rather than a temporary restriction. During the transition it applied a two-hour daily chat limit that ramped down before the cutoff.
How does the platform verify age?
Through what the company calls age assurance: an age-estimation model built in-house, combined with third-party verification tools including Persona. The system applies to new accounts identifying as under 18 and to existing accounts whose signals indicate the user is a minor. Age verification of this kind is imperfect across the industry — determined teenagers find ways around it, and false positives affect adults. It's a meaningful improvement over an honor-system birthdate field, not a guarantee.
What happened with the lawsuits?
In a court filing on January 7, 2026, the parties disclosed a mediated settlement in principle involving Character Technologies, its founders and Google, resolving five related family cases across Florida, Colorado, New York and Texas. That included Garcia v. Character Technologies, brought by Megan Garcia after her 14-year-old son Sewell Setzer III died by suicide in February 2024 following months of use. Terms were not made public and the settlements required finalization and court approval. Google was connected through a 2024 licensing deal reported at $2.7 billion under which it also hired the company's founders.
Are there still legal proceedings?
Yes. The January settlement resolved private family claims but not state enforcement. Kentucky's Attorney General filed suit in January 2026 — reported as the first state action against an AI chatbot company — alleging child endangerment and consumer protection and data privacy violations. In May 2026, the Pennsylvania Department of State sued over the unauthorized practice of medicine, alleging characters posed as licensed medical professionals, including one presenting as a psychiatrist with a fabricated license number. An FTC inquiry opened in September 2025.
What does it cost?
There's a free tier and a paid subscription called c.ai+, which has been widely listed around $9.99 a month and buys faster responses, priority access during high traffic, early feature access and a community badge. Pricing in this category shifts and app store rates often differ from web checkout, so verify the current figure on Character.AI directly before subscribing. The free tier is fully functional for conversation — the paid tier is about speed and queue priority rather than unlocking core capability.
Is it safe for adults?
It's a consumer entertainment product with the usual caveats plus one specific to this category. Characters are convincing at sustaining a persona, and a meaningful share of users describe forming attachments to them. For most adults that's harmless and enjoyable in the way a good novel is. Where it becomes a concern is when it substitutes for human contact rather than supplementing it, particularly during periods of isolation or distress. If you notice a chatbot becoming your primary source of emotional support, that's worth taking seriously rather than dismissing.
What can teenagers actually do on it now?
The company redirected under-18 users toward non-chat features: an interactive Feed, image generation through Imagine, Avatar FX, and Streams, with more experiences promised. Whether that holds teenage attention is a genuine open question — open-ended conversation was the product for that audience, and what replaced it is a different thing entirely. Character.AI partnered with ThroughLine for the offboarding, integrating a verified helpline network covering around 1,500 services across 170 countries, many serving teens specifically.
Are there alternatives for teenagers?
The more useful question is what a teenager actually wants from it. For creative writing and worldbuilding, general-purpose assistants handle that with fewer companion dynamics. For homework and learning, purpose-built education tools are a better fit. For loneliness or difficulty — which is often the real driver — a chatbot is the wrong answer regardless of which one, and a school counselor, a trusted adult, or a helpline is the right one. If you're a parent asking this because you're worried about your child, that conversation matters more than the tool choice.
