Table 2.

Comparison of consistency values within and between models

Group	ChatGPT 5.2	Gemini	DeepSeek	Test statistics	p x
A	100 (100: 100) / 19,5 a	100 (83,33: 100) / 15,55 ab	95,83 (58,33: 100) / 11,45 b	6,912	0,032
B	100 (66,67: 100) / 15,75	100 (66,67: 100) / 16,35	100 (58,33: 100) / 14,4	0,367	0,832
C	100 (58,33: 100) / 13,65	100 (91,67: 100) / 18,35	100 (58,33: 100) / 14,5	2,676	0,262
D	100 (91,67: 100) / 16,65 ab	100 (100: 100) / 18 a	100 (83,33: 100) / 11,85 b	6,412	0,041
E	95,83 (75: 100) / 14,45	100 (50: 100) / 15,75	100 (83,33: 100) / 16,3	0,3	0,861
F	100 (100: 100) / 18 a	100 (100: 100) / 18 a	95,83 (66,67: 100) / 10,5 b	11,484	0,003
Total	100 (58,33: 100) / 94,94 a	100 (50: 100) / 99,93 a	100 (58,33: 100) / 76,63 b	11,015	0,004
Test statistics	13,028	8,260	1,348
p x	0,023	0,143	0,930

^x Kruskal–Wallis H test; median (minimum–maximum)

^a–b No significant difference between language models sharing the same letter. Bold values indicate statistically significant differences (p < 0.05).