Is Character.AI safe for 13-year-olds?
The short answer
Character.AI's own rules allow 13-year-olds (16 in the EU), so the honest question isn't whether they're allowed, it's whether it's wise unsupervised. Our assessment scored it 46 out of 100, which in our rubric means supervision advised.
The risk profile is different from ChatGPT: less about facts and homework, more about open-ended roleplay with characters designed to be engaging, where the concern is attachment and time rather than any single message.
What Character.AI actually is
It's not a homework tool that happens to chat; it's a stage full of characters, many user-created, built for immersive conversation and roleplay. That's why teenagers love it, and why sessions run long. A conversation with a favourite character can continue for weeks and feel like a relationship, because being felt as one is the product working as designed.
Why we scored it 46/100
In our published assessment, Character.AI sits well below the mainstream assistants. Under-18 accounts get a separate model and content restrictions, and the platform added parental-insight features after sustained pressure, including litigation in the US following cases involving teenagers. But moderation of user-created characters is inconsistent, romantic and dependency-shaped roleplay is core to the product, and the age gate is self-declared. Every claim in the assessment carries its source and date; the full profile is linked below.
If your 13-year-old already uses it
A ban tends to move it to a friend's phone rather than end it, so start with interest instead: which characters, what do they like about them, would they show you a bit. Two patterns are worth gently watching for: conversations they're secretive or defensive about in a way that goes beyond normal privacy, and distress when they can't access it. Time-of-day matters too; the 11pm marathon session is where casual use tips into something heavier.
Where Halo Aware fits
Character.AI is one of the twelve platforms we monitor, alongside the other companion and roleplay apps (Janitor AI, PolyBuzz, SpicyChat) that most parental tools ignore. You see topics and patterns, and our alerts specifically watch for emotional-dependency signals, the risk this platform actually carries, rather than only scanning for shocking words.
The companion apps are exactly what we cover
Paid plans monitor all 12 platforms including Character.AI and the roleplay apps most tools ignore. Summaries and dependency-pattern alerts, never transcripts.