The most painful year judging a literary prize: Zhu Youxun says about 40% of submissions were written by AI
Translated from Chinese and summarized by DistantNews. Read the original for the full story.
At a glance
- Taiwanese writer Zhu Youxun says about 40% of the fiction submissions he reviewed this year appeared to be AI-generated.
- He says AI submissions have sharply increased judgesโ workload and damaged the conditions and quality of literary prize reviews.
- Zhu plans to reject judging invitations from next year unless contest rules prohibit AI-created work or require stronger safeguards.
Writer Zhu Youxun has called this his most painful year as a literary prize judge, saying roughly 40% of the hundreds to thousands of fiction submissions he reviewed appeared to be AI-generated.
a certain similar style appeared too many times
In a Facebook post, Zhu said the difference between AI submissions in the first and second halves of the year became increasingly clear. Judges sometimes had to finish an entire batch before recognizing a new AI writing pattern because โa certain similar style appeared too many times.โ Instead of simply selecting the strongest works, he said, judges now also have to determine whether a work was written by a human.
Zhu compared the task to baseball: a pitcher throws three balls at once, but two are illusions, forcing the batter to identify the real one before swinging. He said that if 20 AI-generated works are mixed into a group of 200, the effort feels like reading 400 or 500 submissions. In his view, the current system damages review quality and cannot defend itself effectively.
purely to avoid working myself to death
He argued that judging conditions have collapsed and that fees would need to rise substantially to restore quality. From next year, Zhu said he would first check whether a literary prizeโs rules explicitly ban AI-created work. If they do not, he will refuse the invitation. He said he would give priority to contests requiring contestants to sign declarations or preserve records of their creative process.
Even if my accuracy were as high as 95%, I would still misjudge 50 works out of 1,000.
Zhu also acknowledged the risk of misjudgment. He could label human work as AI-generated or miss AI work, and even 95% accuracy would mean incorrectly assessing 50 submissions out of 1,000. He warned that continued AI submissions could eventually break the prize system, while saying his decision was โpurely to avoid working myself to death,โ rather than a matter of his literary views.
Treasure the last few chances you have to trick prize money out of the system.
Originally published by Liberty Times in Chinese. Translated, summarized, and contextualized automatically by DistantNews, with a note on how the source frames the story. Not individually reviewed before publishing. How this works.