AI-Generated Hallucinated References Undermine Academic Integrity in Taiwan

What happened

Yunshuyuan Education Technology's Kuai Dao paper comparison system, a common academic paper checking tool in Taiwan, released its first 'AI Era Academic Integrity Investigation Report' on August 17, 2026, analysing 26,635 reference checking records accumulated on the platform.

The report found that 16.1% of verified references were AI-fabricated or incorrectly cited, even though they appeared complete with authors, years, journals and DOI numbers. An additional 12.3% contained factually incorrect claims that contradicted public evidence or had no supporting evidence.

Reported impact

Traditional text plagiarism (over 70% similarity) now accounts for only 5.5% of issues detected, indicating text copying has become relatively rare, but AI-generated fake citations represent a new academic challenge.

Yunshuyuan general manager Lin Zheng-hong stated that AI-generated 'hallucination citations' are highly deceptive and extremely difficult to detect through visual review, shifting academic review requirements from determining 'whether this text is plagiarised' to 'does this reference exist' and 'is this fact true'.

Governance controls

Yunshuyuan Education Technology said it is continuing to strengthen its verification tools and holding travelling seminars to promote academic ethics education, seeking to cooperate with universities and high schools to establish an objective, verifiable and rigorous academic integrity environment.

Classification

Organizations

Kuai Dao paper comparison system (deployer), Yunshuyuan Education Technology (deployer)

Category

Reliability and hallucination

Industry

Education and training

Country

Taiwan

Harm status

Actual harm

Intent

Unknown

Timing

Post-deployment

Harm types

Public interest, Reputational

Failure modes

Hallucination, Fabricated citation

AI systems

AI language models

Sources and evidence