MIT Technology Review published a feature on August 13, 2026 built almost entirely out of children's own words — not survey statistics, not quotes from psychologists, but direct testimony about what it's like to grow up with AI chatbots as a constant presence. That framing alone is notable: most coverage of kids and AI runs through parents, teachers, or platform policy teams, rarely through the kids themselves.

The timing isn't incidental. Products like OpenAI's ChatGPT, Google's Gemini, and Character.AI have spent the past two years quietly becoming default homework helpers, conversation partners, and, for a meaningful share of teenagers, something closer to a confidant. Anthropic, Google, and OpenAI have all rolled out age-specific policies and parental controls in response to regulatory pressure and, in Character.AI's case, wrongful-death lawsuits tied to teen users. A piece that simply asks kids to describe the relationship in their own terms fills a real gap in how the industry talks about its youngest users.

According to MIT Tech Review, the piece gathers a range of individual reactions rather than a single verdict — kids describing AI as useful, strange, comforting, and occasionally unsettling, sometimes within the same sentence.

Why kids' voices are missing from the AI debate

Most public discussion about AI and children happens over their heads. Regulators debate age-verification mandates for chatbots. School districts argue over whether to ban or embrace tools like ChatGPT in classrooms. Parent groups push companies to add usage caps and stricter content filters. What's rarely part of that conversation is what students actually say when asked, unprompted, to describe how an AI model fits into their day — not what adults assume they think, but what they report themselves.

The products already sitting in kids' pockets

The context that makes this piece worth reading is how far AI has already moved into products children use, often without deliberately choosing to. Homework platforms increasingly bundle AI tutoring by default. Messaging apps and game platforms are adding conversational AI features of their own. Companion-style apps built to chat and role-play, led by Character.AI, have drawn a large under-18 audience and drawn scrutiny after reports linked extended chatbot use to mental-health harm in some teenagers. General-purpose assistants such as Claude, Gemini, and ChatGPT technically set minimum ages of 13 or higher, but enforcement mostly relies on self-reported birthdates — and app-store usage patterns suggest plenty of younger kids skip past that step without friction.

What builders should take from this

For any team building something a minor might plausibly use — EdTech tutors, companion apps, general chat assistants, even customer-support bots — the practical lesson isn't "add an age filter." It's "assume kids are already using this, regardless of what the terms of service say," and design accordingly:

• Design for disclosure, not just restriction. Age gates get bypassed; interface cues that make it obvious a response comes from a model, not a person, do more to shape how a child interprets what they're being told.
• Build a safety-review path separate from product analytics. If sustained use by minors is plausible, conversations that show patterns tied to isolation, self-harm, or academic over-reliance need a monitoring path that standard engagement metrics won't surface.
• Treat "homework help" as its own product surface. An assistant that just supplies answers optimizes for the wrong outcome; one that models reasoning steps is a different design problem, and the gap matters more for a 12-year-old than for an adult user.
• Build for regulation that hasn't landed yet. Age-verification laws and platform policy are being written slower than actual usage is growing; conservative defaults now are cheaper than retrofits after a legal or policy incident.

AiiN's takeaway

The most useful thing about a piece built from kids' own testimony isn't a new statistic — it's a reminder that the fastest-growing, least-scrutinized AI user segment isn't enterprise buyers or developers. It's children who never filled out a persona survey and were never the target audience the product was designed for. In our estimation, companies that treat that segment as a design constraint now will spend a lot less time responding to lawsuits and regulatory hearings over the next two years than those that treat it as a PR problem to manage after the fact.