Dutch thesis analysis flags 20% as predominantly AI-written

An analysis of recently completed master’s theses from 12 Dutch universities classified about one in five as containing predominantly artificial intelligence-generated content, according to an investigation reported on Wednesday 14 October. The analysis used Pangram, software designed to identify text potentially generated by chatbots.

The investigation found that 6% of the theses were classified as almost entirely AI-generated and 14% as containing more than 50% suspected AI-written content. Another 39% contained smaller amounts of flagged text, while 41% showed no detectable AI-generated content.

The investigation said automated detection is not conclusive proof that a chatbot produced a passage and requires further examination before a university can determine whether academic rules were broken.

Dutch Education Minister Rianne Letschert called using AI to write a student’s thesis academic fraud. She said universities should strengthen oversight and said a strategy addressing the issue would be introduced before the end of 2026.

The investigation also found that the 12 universities and Universities of the Netherlands did not keep comprehensive figures on AI-related thesis fraud. It said policies on generative AI varied between universities and, in some cases, between courses.