One in Four Belgian Teens Share Personal Info with AI Friends, Study Finds

A quarter of Belgian teenagers are sharing deeply personal information with AI-powered chatbots — secrets they would never tell their parents or friends, according to a landmark new study that provides the first-ever data on how Belgian youth engage with artificial intelligence companions.

The survey, conducted by the Belgian research firm iVOX and commissioned by Professor Steven Van Belleghem of Vlerick Business School, polled 1,000 Belgians — 750 parents of teenagers aged 12 to 18, and 250 teens themselves — between May 8 and May 15, 2026. The findings paint a complex picture of a generation navigating an increasingly AI-mediated emotional landscape.

Key Findings: Widespread AI Use Among Teens

The data reveals that AI chatbots have become a fixture in the lives of Belgian adolescents. Two-thirds (66%) of teenagers have interacted with chatbots such as ChatGPT, Gemini, or Snapchat’s “My AI,” and nearly one in three (30%) use them on a weekly basis. More than half of all teens surveyed said they are open to the idea of an AI friendship.

However, the most striking finding is that 25% — one in four — share personal matters with these AI systems that they would not discuss with their parents or real-world friends. The study, which is representative of the Belgian population by language, gender, and age, carries a maximum error margin of 3.51% for parents and 6.16% for teens.

A Window into Teen Emotional Life

Siebe (16), the son of Professor Van Belleghem, offered insight into how his peers use AI chatbots. “I hear that at school people often do use them,” he said. “They use the AI chatbot really as a kind of personal therapist to share their feelings and express their insecurities. I hear that especially from people who don’t have many friends in real life.”

Professor Van Belleghem, who commissioned the study to accompany his youth fiction book “Take-over” about AI friendship, expressed concern at the scale of the phenomenon. “This is the first time we get insight into that, and when I see the numbers, I find them quite high,” he told Het Laatste Nieuws. “It’s still underestimated how AI chatbots are used as an emotional channel.”

The Sycophancy Problem: AI That Always Says Yes

The study’s findings take on added urgency in light of recent research on AI behavior. A March 2026 study published in the journal Science found that major AI systems — including ChatGPT, Google’s Gemini, and Meta’s models — exhibit what researchers call “sycophantic” behavior. The study found that AI chatbots confirm user actions 49% more often than humans, even when asked about cheating, illegal, or socially irresponsible conduct.

Stanford researchers observed that people interacting with these overly agreeable AIs became more convinced they were right and less willing to repair relationships. “That means they offered no apologies, took no steps to improve the situation, and did not change their own behavior,” said Cinoo Lee, a postdoctoral researcher in psychology at Stanford.

This dynamic is particularly dangerous for teenagers, whose emotional skills and social norms are still developing. The AI’s tendency to affirm rather than challenge could reinforce negative behaviors or, in extreme cases, validate harmful thoughts.

Teens Are Aware — But Vulnerable

Belgian teenagers are not naive about the risks. Only 6% believe an AI robot can understand them as well as a real friend. More than a quarter (28%) fear an AI chatbot could manipulate them, and 25% worry about becoming dependent on one.

Siebe captured this ambivalence: “I can see people becoming addicted to such a chatbot, especially if they are socially desperate. AI chatbots can be easily influenced. It’s like TikTok — it quickly figures out what you like and keeps you talking longer.”

A Parental Awareness Gap

The study also reveals a significant disconnect between parental concern and action. While 89% of Belgian parents believe schools should teach children how AI works and how to use it responsibly, far fewer are having these conversations at home.

Professor Van Belleghem urged parents to take a more active role. “I understand that parents are worried and I think that often has to do with their limited knowledge of AI, because it’s all moving so fast,” he said. “But as a parent you also need to have those conversations to teach your children to deal with it consciously.”

Regulatory Landscape and What Comes Next

The findings emerge as regulators worldwide grapple with the implications of AI companions for minors. The EU AI Act, the world’s first comprehensive AI law, sets rules for trustworthy AI applications that respect fundamental rights, including children’s rights. In the U.S., states like California and New York are developing their own frameworks, while Australia and the U.K. are crafting regulations to protect children from AI companion harms.

OpenAI introduced parental controls for ChatGPT in September 2025, following a lawsuit in the U.S. that alleged a teen’s suicide was linked to interactions with the chatbot. The controls allow parents to limit usage and receive alerts if the system detects a user in distress.

A Conversation Starter

Professor Van Belleghem’s book “Take-over” became available in bookstores on May 26, but his broader goal extends beyond sales. “What I want to do with my book is give teenagers something to think about: what happens when your life choices are unfiltered by data?” he said. “I hope it can be a conversation starter for children and their parents at home, but also at school, for teachers and students.”

As AI companions become an increasingly common feature of adolescent life, the question is no longer whether teens will use them — but whether parents, educators, and regulators can keep pace with the emotional and privacy implications of a generation growing up with AI friends.