Anxious Chinese students are trusting AI to help pick colleges and majors

Guo Xinyan had three weeks after China's brutal gaokao exam to decide her entire future, so she typed her score, her province ranking, and a vague wish list — law, accounting, finance, a coastal city, moderate tuition — into Alibaba's Qwen chatbot. It spat back dozens of options. Her parents, meanwhile, asked ByteDance's Doubao to do the same thing. Neither Guo nor her family had ever navigated a college application before, and in a system where roughly 10 million students take this single exam every year and get only weeks to lock in their choices, that kind of guesswork used to cost real money.

That's exactly the gap China's biggest tech companies are racing to fill. Since this year, Alibaba, ByteDance, Tencent, and Baidu have all rolled out agentic AI tools built specifically for college and major selection, feeding on historical admission cutoffs, personality tests, and student preferences to sort programs into reach, safety, and backup tiers. Alibaba says it brought in 300 college-search specialists to sharpen Qwen's reasoning and had generated 23 million recommendation reports by July 8. Tencent's Yuanbao claims 200 million admission-related queries answered by July 10. Baidu says 15 million people used its advisory feature by late June, backed by human review of the AI's picks. This isn't a side feature — it's a full-on land grab for daily chatbot habits, dressed up as public service.

Before any of this, families paid for the service instead, sometimes handing over thousands of dollars to consultants like Zhang Xuefeng, a Douyin personality with more than 27 million followers who built a brand on blunt, money-first advice — pick engineering, skip philosophy — and charged up to $2,700 to help fill out forms. Zhang died suddenly in March. Within weeks, someone had turned his books and sayings into a downloadable 'skill' on GitHub, letting anyone spin up a chatbot that talks like him. One user, designer Xindy Lin, stitched that skill together with admissions data and a program guide on OpenAI's Codex to help her sister find majors with solid job prospects that didn't demand heavy math. It worked, more or less — the bot suggested things like supply-chain management.

But cheap access to sorting power isn't the same as good judgment, and that's the part nobody's automated yet. Ye Xiaoyang, an economist who advises rural students, points out that plenty of families still don't know how to prompt a chatbot well enough to get useful answers, and that AI is good at filtering thousands of programs but useless at telling a teenager what to do with their life. Chinese media testing has found these tools citing outdated figures or leaving strong options off the list entirely. Worse, because students with similar scores tend to get similar recommendations, Ye warns the bots risk funneling whole cohorts toward the same handful of programs — and missing a match in this one-shot system can mean landing somewhere far below where a student could have gotten in.

Guo ended up doing what a lot of students still do: she used the AI, then didn't fully trust it. She checked Qwen's suggestions against official guidebooks and against what actual college students were posting on RedNote about dorm conditions and campus life. Her reasoning was simple. Real people's experiences, she said, still felt more trustworthy than a report generated in seconds.