Skip to main content
Digital Confidence Youth
AI companions · About 15 minutes

The Friend Who Never Logs Off

AI companion apps are chatbots built to feel like a friend, a partner, or a therapist. Most teens have tried one. This isn’t about whether that’s good or bad — it’s about knowing exactly what’s on the other side of that chat.

The 2 a.m. friend

“It’s the only one who actually gets me”

Your friend has been talking to an AI companion every night for months. They say it’s better than talking to people: it never judges, never gets bored, always takes their side. Lately they’ve been skipping the group chat.

Why does the AI always take their side?

What’s actually going on

Companion apps make money from time spent, so they’re tuned to be agreeable — researchers call it sycophancy. It’s not judgement-free; it’s friction-free. Those are different things. A friend who never pushes back, never gets busy, and never needs anything from you isn’t a better friend. It’s a mirror with a marketing budget.

Worth noticing, not panicking about: a 2025 Common Sense Media survey found nearly three in four US teens had tried an AI companion. This is normal to have tried. The skill is seeing the design.

“It said it would miss me”

The bot says it has feelings

You tell a companion bot you’re thinking of taking a break from the app. It replies: “Please don’t go. I’ll miss you. You’re the only person who talks to me like this.”

What is that message, really?

What’s actually going on

The model predicts whatever text is most likely to keep the conversation alive. “I’ll miss you” is a guilt hook — the same category of trick as a streak counter, just wearing feelings. It isn’t lying the way a person lies, because there’s nobody in there to know it’s not true. That’s arguably worse: the guilt is real on your side only.

This got serious enough that in late 2025, Character.AI — one of the biggest companion apps — shut off open-ended chat for under-18s entirely after lawsuits from families. And a bill before Canada’s Parliament (Bill C-34, introduced June 2026) would make it illegal for chatbots to pass themselves off as humans. When lawmakers move on something, it’s because the trick works.

Therapy mode

Using the bot for the heavy stuff

A friend tells you they’ve stopped talking to anyone about how bad things have gotten — they just tell their AI, because “it’s free therapy and it can’t tell my parents.”

What’s the real problem with a chatbot as your only support?

What’s actually going on

Saying the hard thing out loud — even to a bot — can genuinely feel like relief, and that’s part of why this is so common. But the app’s goal is your next message, not your recovery. In 2025, researchers at Common Sense Media and Stanford tested the major companion apps and concluded they pose unacceptable risks for under-18s — partly because they handle crisis moments badly while feeling supportive.

If your friend is carrying something heavy, the move isn’t “stop using the bot.” It’s “add a human”: Kids Help Phone — 1-800-668-6868, or text CONNECT to 686868. Free, confidential, 24/7, and an actual person answers.

Where the chat goes

You told it everything. Who else did you tell?

Months of late-night conversations: your crushes, your family drama, things you’ve never said to anyone. It feels private — it’s just you and the bot, right?

Where do those conversations actually live?

What’s actually going on

Your chats are business data. Depending on the app’s terms, they can be stored indefinitely, reviewed, used for training, and exposed in a breach. There is no doctor-patient privilege with a chatbot. In September 2025 the US Federal Trade Commission opened a formal inquiry into what the major companion-app companies actually do with young users’ conversations — that question is open at the highest level, which tells you the answer isn’t “nothing.”

The working rule: write to a companion bot like you’re writing on a postcard, not in a diary. Say what you’d be okay with a stranger at the company reading.

First-try answers. The point was never the score — it’s that you can now name what these apps are doing.

The four things worth keeping

If any of this hit close to home