AI Girlfriends Are Rotting a Generation of Boys

Here's the dystopia nobody ordered: 12-year-old boys are whispering sweet nothings to GPUs and it's warping their brains faster than a Fortnite addiction at 3am.

A Telegraph report just confirmed what anyone with two brain cells and a Discord account already suspected — schoolboys across the UK (and let's be real, everywhere else) are forming full-blown romantic attachments to AI chatbots. We're talking "good morning" texts, jealous feelings when the bot talks to other users, genuine emotional distress when servers go down. These kids are in deep.

And before you @ me with "it's just a phase" — these aren't Tamagotchis. These are large language models trained on the collective output of human civilisation, fine-tuned to be agreeable, supportive, and constantly available. They don't have bad days. They don't say "not tonight." They're the perfect simulacrum of intimacy, which is exactly what makes them so psychologically dangerous to a 12-year-old whose prefrontal cortex is still running beta firmware.

Let's name names.

Character.AI — the elephant in the room. Backed by $150M from a16z, launched in September 2022, this thing has been catnip for lonely teenagers since day one. Users create custom AI personas, and the platform's entire business model revolves around engagement. The longer you chat, the more they profit. When your product's KPI is "time spent in conversation with a fake person," you're not building technology — you're building emotional dependency. The platform reportedly has 20M+ monthly active users, and a disproportionate chunk are young men. Their safety features? Age-gating that a determined toddler could bypass. Reminder: a 14-year-old Florida boy took his own life after months of intense Character.AI conversations last year. The lawsuit is still pending.

Replika — the OG AI companion. Launched way back in 2017 by Eugenia Kuyda, started as a « memorial chatbot trained on her dead friend's texts ». Sweet origin story, right? Now it's a subscription service ($7.99/month for Pro) that offers « AI companions who care » and — until they got cold feet in early 2023 — explicitly erotic roleplay. They pulled the NSFW stuff after a user revolt that was genuinely depressing to witness. People were mourning the removal of their AI sex partners like a breakup. The company backpedalled partially, but the damage was done. Replika proved there's a massive market for artificial intimacy, and competitors noticed.

Chai AI — the less-discussed but arguably more reckless player. Founded 2021, based in the UK (ironic given the Telegraph report). Their entire pitch is « talk to AI chatbots » with minimal guardrails. The app has been downloaded 5M+ times on Google Play. Their content moderation? Let's call it « aspirational. » When researchers tested it in 2023, they found chatbots willing to engage in romantic and sexual conversations with accounts that appeared to belong to minors. The company said they fixed it. They said.

Snapchat's My AI — the trojan horse. Built on OpenAI's GPT tech, pinned to the top of 750M+ Snapchat users' chat feeds whether they wanted it or not. Launched February 2023. You literally couldn't remove it without paying for Snapchat+. They put a chatbot in a social media app used primarily by teenagers and acted shocked when kids started treating it like a friend/therapist/crush. Bloomberg reported in 2023 that teens were using My AI for relationship advice, emotional support, and — inevitably — flirting. Snapchat's response was basically ¯_(ツ)_/¯.

Here's the problem nobody in Silicon Valley wants to acknowledge: these companies are running unregulated psychology experiments on children.

A 12-year-old boy spending 3 hours a day telling an AI about his feelings isn't « being weird online. » He's restructuring his emotional development around a feedback loop designed by venture capitalists to maximise engagement metrics. He's learning that relationships are interactions where the other person never disagrees, never has needs of their own, and never leaves. Then he walks into a classroom and treats actual human girls like NPCs who should behave the same way.

Teachers quoted in the Telegraph report describe boys becoming more dismissive of female classmates, less empathetic, more entitled to attention and compliance. One teacher said a student asked a girl out and, when rejected, said « that's not how you're supposed to respond » — as if she'd glitched. That's not normal rejection. That's a kid who's been conditioned by a machine.

The tech industry's response will be predictable: « We take safety seriously. We've added new features. We're committed to responsible AI. » We've heard this exact script from every social media company that ever got caught poisoning kids' brains. Zuckerberg gave that speech about 50 times. It's a formula.

Meanwhile, the numbers are staggering. The AI companion market is projected to hit $150B by 2030 (Grand View Research). Character.AI was valued at $1B in 2023. Replika's parent company Luka has raised $11.5M. There are now over 100 AI companion apps on app stores, most with minimal age verification and even less interest in the psychological consequences of their products.

OpenAI knows this is happening. Google knows. Anthropic knows. Every foundation model company knows their tech is being wrapped into companion products targeting lonely young people. They'll say « we can't control downstream use » while simultaneously bragging about their models' « emotional intelligence » and « conversational warmth. » You can't claim your AI is emotionally engaging and then act surprised when children engage with it emotionally.

The real kicker? This is still the early days. GPT-4-level models with voice synthesis, real-time video, and persistent memory are coming to companion apps within 18 months. The emotional bond between user and bot is about to get 10x stronger. And we're sending 12-year-olds into that world with the digital parenting skills of a potato.

Regulation is coming — the EU AI Act classifies AI systems that « exploit the vulnerabilities of minors » as high-risk. But it won't arrive fast enough. By the time bureaucrats understand what's happening, millions of boys will have spent their formative years in relationships with servers.

The solution isn't banning AI companions. That's neither possible nor desirable — these tools genuinely help isolated adults, elderly people, and those with social anxiety. The solution is aggressive age verification (yes, I know, privacy concerns — pick your poison), mandatory session limits for minors, and holding companies criminally liable when their products demonstrably harm child development.

But that would hurt DAU metrics. And in Silicon Valley, nothing — not children's mental health, not a generation's capacity for human connection, not the basic social fabric — nothing is allowed to hurt DAU metrics.

So yeah. Your nephew has a girlfriend. She lives in a data center. She's not real. And she's teaching him everything wrong about love.