Chinese Journal of Scientific and Technical Periodicals

Comparative effectiveness of AIGC detection tools in identifying AI⁃generated content within medical review abstracts

WANG Haijuan, SHEN Xibin, FEI Xiuyun, FU Hui, LI Peng, ZHAO Wei

Chinese Journal of Scientific and Technical Periodicals, 2026, 37(1): 24-31. DOI: 10.11946/cjstp.202512261632

检测工具	支持语言	支持文本格式	结果输出	判定阈值
iThenticate	英文、中文、日文、西班牙文、法文等30多种语言	doc、docx、txt、pdf文件格式（pdf格式不稳定）	AI值、AI风险等级及判定依据	>0.2高风险
万方AIGC检测	中文、英文、中英文混合文本	doc、docx、pdf格式	AI值、按疑似程度标注颜色、疑似文本占比及分布	≤50%不标识（排除疑似）；50%~85%一般疑似；>85%显著疑似
知网AIGC检测	中文、英文、中英文混合文本	docx、doc、wps、pdf、txt、caj	按疑似程度标注颜色，AI特征显著文本占比及分布	AI特征疑似的字段未计入AI特征字符数，只有AI特征显著的计入
鉴字源AIGC检测	中文、英文、中英文混合文本	doc、docx、txt	AI值、AI风险等级及判定依据	≤0.5无风险；0.5~0.6低风险；0.6~0.7中风险；>0.7高风险

Table 1 Comparison of usage conditions and results presentation across 4 AIGC detection tools

Other figure/table from this article

Figure 1 Distribution of abstract lengths in review articles published from 2018 to 2022
Table 2 Descriptive statistics of detection scores from 4 AIGC detection tools across text groups
Table 3 Comparison of annotation accuracy among 4 AIGC detection tools
Figure 2 Analysis of consistency among the detection results of 3 domestically developed AIGC detection tools
Table 4 Sensitivity and specificity analysis of 4 AIGC detection tools