. 2024 Jun 26;26:e52001. doi: 10.2196/52001

Table 1.

Baseline data in this study.

Variables	Human (n=60), mean (SD)	ChatGPT (n=60), mean (SD)	Bard (n=54), mean (SD)	P value	95% CI
Similarity index (%)	100 (0)	20.7 (8.7)	32.1 (11.0)	<.001	–15.05 to –7.62
AI^a detection rate (%)	28.4 (25.8)	60.7 (25.0)	77.7 (21.1)	<.001^b	N/A^c
Word count (n)	317.8 (72.4)	317.8 (53.9)	223.8 (44.5)	<.001^d	N/A

^aAI: artificial intelligence.

^bBonferroni post hoc analysis results for the AI detection rate were as follows: human versus ChatGPT: P<.001; 95% CI –42.92 to –21.58; human versus Bard: P<.001; 95% CI –60.18 to –38.26; and ChatGPT versus Bard: P=.001; 95% CI –27.93 to –6.01.

^cN/A: not applicable.

^dFor text number count, Bonferroni post hoc analysis showed the following: human versus ChatGPT: P>.99; 95% CI –25.89 to –25.29; human versus Bard: P<.001; 95% CI 67.41-120.49; and ChatGPT versus Bard: P<.001; 95% CI 67.46-120.54.