evidoria

← Back to browse

Good practice Imported

AI Gaokao Application Agents — China's Tech Platforms Guide Millions to University Places

China · Beijing · See the China profile · See the Beijing profile

Evidence: Descriptive / self-reported Top 82% 33/100 · Ask Evidence Copilot about this practice

In June 2026, Alibaba, Tencent and Baidu launched free AI agents that turn gaokao scores into ranked university and major choices. Within three weeks, ~14-15 million users each tried Alibaba's and Baidu's tools, and Tencent's bot fielded about 80 million questions.

12.9 million
Students who sat the 2026 gaokao (2026)
14 million
Users of Alibaba's admissions-support function (June 2026)
15 million
Users of Baidu's admissions-support feature (June 2026)
80 million
Admissions-related questions fielded by Tencent's Yuanbao (June 2026)
78.4 %
Candidates who used AI to search universities or generate application plans (2026)
92 million
Projected national users of AI college-application tools (2026 (projected))
AI Gaokao Application Agents — China's Tech Platforms Guide Millions to University Places

Details

Maturity
Scaling
Promoter
Alibaba (Tongyi Qianwen/Qwen), Tencent (Yuanbao) & Baidu — national AI admissions-guidance agents
Period
June 2026 – present
Keywords
generative AI, higher-education admissions, career guidance, edtech

Context

China's gaokao is a single national exam that determines university access for secondary-school leavers; choosing among nearly 3,000 universities and 2,000-plus majors historically required manual research or a paid admissions-consulting industry affordable to fewer than 5% of families.

Objectives

Give students free, instant AI-generated guidance on which universities and majors to apply to based on their exam score and provincial rank.

Activities

On 8 June 2026, as the exam period opened, Alibaba (Tongyi Qianwen/Qwen), Tencent (Yuanbao) and Baidu each launched AI agents generating ranked 'ambitious', 'matched' and 'safe' university-major lists from a student's score and rank; Alibaba's tool drew on eight years of exam-guidance data from its Quark assistant.

Results

By late June 2026, about 14 million people had used Alibaba's function, about 15 million had used Baidu's equivalent, and Tencent's Yuanbao had fielded around 80 million admissions-related questions, against 12.9 million gaokao test-takers; state media reported 78.4% of candidates used AI to search for universities or generate application plans, with industry trackers projecting 92 million national users of such tools in 2026. Chinese outlets including Xinhua and Global Times also reported AI hallucinations, including fabricated programmes and majors.

Conclusions

Adoption at this scale suggests genuine reach, but all figures are self-reported by the platforms themselves, no independent accuracy audit has been published, and education experts quoted in Chinese media caution that AI accuracy 'still falls short' of careful manual research.

Implementation

Indicative cost
High (€500k–€5M)
Time to results
Short (< 1 year)
Staffing & skills
Alibaba (Tongyi Qianwen/Qwen) engineering team, Tencent (Yuanbao) team, Baidu team

Conditions for success

  • Existing large consumer AI assistants with prior exam-guidance data (e.g. Alibaba's eight years of Quark data)
  • Free access removing the cost barrier of paid admissions consulting
  • Timing aligned precisely with the exam and application window

Common failure modes

  • AI hallucinations producing fabricated programmes/majors, reported by Xinhua and Global Times
  • No independent audit of recommendation accuracy
  • Adoption figures are self-reported by vendors with a commercial incentive to inflate usage

Where it fits

Governance type
private-sector platforms, no government mandate documented
Scale
national (12.9 million exam takers)
Income level
upper-middle-income (China)

Do you run this practice? Claim it — verified implementers get a public contact pathway and can propose corrections.

Data sources

Where this practice's information was retrieved from, and when.

Attachments

Similar practices you may find useful