China's Students Bet Their Futures on Free AI College Advisers

The Three-Week Window

After enduring China's national college entrance exam, 18-year-old Guo Xinyan had exactly three weeks to choose her future. She opened Alibaba's Qwen chatbot and fed it her score, her provincial rank in Shandong, three acceptable majors, a modest budget, and a preference for coastal cities. Within moments, the system returned dozens of university-program pairings sorted by probability of admission.

Her parents, unfamiliar with the application process, consulted ByteDance's Doubao for a second opinion. Neither family member mentioned passion, intellectual curiosity, or long-term learning goals. "There's nothing in particular I want to study," Guo explained. "We are just looking for good career prospects."

At DailyTechWire, we've tracked the rise of agentic AI in consumer decision-making across Asia, but few applications carry stakes as high as this one. Around 10 million Chinese students sit the gaokao each year, and the compressed timeline between score release and application deadlines leaves little room for deliberation. Unlike U.S. undergraduates who routinely switch majors, Chinese university policies make transferring between disciplines difficult and rare. For most students, the choice made in those three weeks is final.

A $160 Million Industry Under Pressure

Before 2025, families navigated the labyrinth of thousands of schools and hundreds of majors through a patchwork of guidance books, social media influencers, and private tutors. The market for college admission advisory services reached $160 million, according to consultancy iiMedia. Premium one-on-one coaching sessions commanded fees as high as $2,700.

The most prominent figure in this space was Zhang Xuefeng, a blunt-spoken adviser with over 27 million Douyin followers. Zhang built his reputation by telling working-class families which fields led to stable employment and which did not. Math and engineering earned his endorsement; journalism and philosophy did not. "If your family has no money, don't talk about dreams," he would say. "Pick a major that puts food on the table."

Zhang died suddenly from cardiac arrest in March, but his influence persists. Developers extracted his books and recorded remarks to create a "Zhang Xuefeng skill" on GitHub, a bundle of instructions that can be loaded into any AI agent to replicate his utilitarian tone. The code is open-source and widely used.

The Big Four Enter the Market

Alibaba, ByteDance, Tencent, and Baidu all launched specialized college advisory chatbots in 2025, offering the service at no charge. The tools are built on agentic AI architectures that analyze historical admission cutoff scores, ask for gaokao results, career interests, and even Myers-Briggs personality types, then produce tiered recommendations labeled "reach," "safety," and "backup."

Alibaba claims Qwen generated 23 million college program reports for this year's gaokao cohort and has conducted tutorials in rural high schools. Tencent's Yuanbao bot reported fielding 200 million admission-related queries by mid-July. Baidu said 15 million people used its advisory feature in late June alone.

The companies frame the initiative as a public service, but the strategic motive is clear: habituating millions of young users to rely on AI for high-stakes decisions. If a chatbot earns trust during the most consequential choice of adolescence, it stands a better chance of becoming the default tool for career moves, financial planning, and major purchases later.

Alibaba has staffed Qwen's college advisory function with 300 human specialists to refine the model's reasoning. Baidu promises that human experts review AI-generated recommendations before delivery. Both measures acknowledge the liability exposure inherent in algorithmic guidance at this scale.

Utility Over Identity

Xindy Lin, a user experience designer in Fujian province, built a custom chatbot on OpenAI's Codex to help her younger sister. She combined the Zhang Xuefeng skill, a directory of available programs, and three years of admission data. Her sister's requirements: good job prospects, minimal math. The bot suggested supply-chain management and e-commerce marketing.

"There's no way individuals can digest so much information in such a short time," Lin said. The logic is defensible. With thousands of permutations and a narrow decision window, algorithmic filtering appears rational. But the framing of the query itself reflects a broader shift in how young people are encouraged to think about higher education.

Career counseling in Chinese high schools is minimal. In an environment of high youth unemployment and widespread anxiety about AI displacing jobs, the calculus has tilted sharply toward instrumental outcomes. Universities have responded by introducing majors in "embodied intelligence" and "brain-computer science." AI itself has become a popular choice among students hedging against automation.

The result is a feedback loop: economic insecurity drives students toward pragmatic majors, which in turn reinforces the perception that education is primarily a credentialing exercise. Chatbots trained on historical admission data and labor market signals amplify this dynamic. They optimize for employability, not for intellectual development or long-term adaptability.

The Hidden Costs of Automation

Ye Xiaoyang, a Seattle-based economist and visiting professor at East China Normal University, has spent years advising rural students on college applications. He acknowledges that AI excels at sorting through tens of thousands of program options, but he remains skeptical of its capacity to help young people determine what they want from life.

"Everyone can open a chat box, but it doesn't mean everyone has the same ability to make choices," Ye noted. Many rural students lack basic computer literacy and struggle to craft prompts that yield useful guidance. This year, Ye launched his own free AI platform designed to teach students foundational admission rules and encourage exploration of career directions rather than simply delivering ranked lists.

"A good AI system should be able to guide students to think for themselves," he said.

The technical limitations are also material. Chinese media testing has uncovered instances of chatbots citing outdated admission cutoffs or omitting strong-fit programs. If a student relies on faulty data and fails to secure a match in the first round, they cascade into lower-ranked schools or miss the cycle entirely. The consequences are not easily reversed.

There is also the herding problem. If millions of students with similar scores receive similar recommendations, the bots may inadvertently concentrate applications around a narrow set of programs, distorting admit rates and creating unexpected competition in previously overlooked fields.

Trust, Verification, and the Human Anchor

Guo, the Shandong student, knows classmates who paid for human coaches. She chose to cross-reference the chatbot's output with official guidance books and posts from current students on social network RedNote. She wanted to know about dormitory conditions, campus culture, and whether the schools delivered on their promises.

"I still find real people's experiences more trustworthy," she said.

Her approach suggests a pragmatic middle path: use AI to narrow the field, then rely on human testimony and institutional track records to make the final call. It is a form of hybrid decision-making that may become standard as algorithmic advice proliferates in domains once reserved for human judgment.

The college admission coaching industry has long been criticized for monetizing anxiety and, in some cases, for coaches secretly using AI to generate recommendations while charging premium fees. But the shift to free, automated tools does not eliminate the underlying problem. It simply redistributes risk from financial cost to informational accuracy and interpretive skill.

For China's tech giants, the gaokao advisory push is both a brand-building exercise and a test of agentic AI's ability to handle nuanced, multi-variable decisions at scale. For the 10 million students who take the exam each year, it is a high-wire act with limited safety nets. The bots are fast, comprehensive, and free. Whether they are wise remains an open question.