The Paper's test of three major AI detection platforms shows wildly inconsistent results, with original papers being misidentified
On November 19, The Paper reported that reporters tested three major AI detection platforms—CNKI, Wanfang, and VIP—and found that the detection rate for the same AI-generated paper varied from 7.15% to 69.8%, while an original human-written paper was flagged as containing 25% to 36% AI content. Some universities use AI detection rates as a standard for graduation review; experts have criticized the algorithms as unreliable and are calling for the establishment of manual review mechanisms.