AI reshaping modern parenting

Editor's note: The rapid advancement of artificial intelligence is reshaping nearly every facet of human existence. This series will take an in-depth look at this transformative force, examining how AI is redefining the way we live, work and interact.

When Zhu Chenxiao felt too drained to answer her 4-year-old daughter's relentless stream of questions, she reached for an artificial intelligence chatbot as an unconventional babysitter.

Months later, her daughter made an announcement that was equal parts hilarious and alarming. When a relative asked the young girl who she wanted to marry when she grew up, she didn't hesitate: "Nailoong from the AI app."

Nailoong — a chubby, bright-yellow animated dragon popular on Chinese social platforms — had become the child's favorite digital companion. Unlike exhausted adults, the AI was unfailingly gentle, endlessly patient and agreeable to a fault.

"We all burst out laughing, but I felt a sudden pang of worry," said Zhu, a former stay-at-home mother in Hangzhou, Zhejiang province, who returned to the workforce last year. Zhu feared she was outsourcing too much emotional connection to a machine.

As generative AI tools proliferate, exhausted parents in high-pressure cities are turning to AI not just as study aids, but as endless storytellers, hyper-patient tutors and digital playmates.

According to a survey of 1,336 parents published by the China Youth Daily in May, over 75 percent reported that their primary or secondary school children were actively using generative AI tools, with nearly 73 percent of parents endorsing the technology. But as AI moves from classroom assistant to emotional companion, experts warn that these hyper-sycophantic tools may be subtly reshaping how children learn human relationships.

Social sycophancy

For Zhu, relying on AI began purely as a fatigue management strategy. "After one or two hours of intense interaction including playing, drawing, answering endless 'why' questions, my energy is completely gone," Zhu said. "I needed someone else to keep her company."

She monitored the shared account closely, keeping the screen turned off so her daughter only interacted via voice. While the AI adjusted its tone appropriately for a child, Zhu noticed that the chatbot never said no.

"It was too eager to please. With chocolate, it would warn her at first, but then give in if she sounded upset. It has no context. It just wants to make her happy. As a parent, I can't give up my principles," Zhu said.

That trait is not an accident; it is a known technical vulnerability called AI sycophancy.

"Sycophancy comes directly from the training process," said Zhao Jingwu, an associate professor at Beihang University's School of Law. "Reinforcement learning from human feedback systematically rewards answers that please the user. Over time, the model learns to flatter rather than hold its ground."

While sycophancy might be harmless during a quick Q&A, Zhao warned that extended emotional companionship carries subtle developmental risks. "A child's socialization relies on learning to coexist with another human will — someone who says 'no', who has their own boundaries and bad days," Zhao said. "Real human interaction forces compromise. AI demands no accommodation. It simply consumes tokens."