Table 2 Frequency of distribution for each ChatGPT attempt. In attempt 1, 80% (n = 40) of answers were accurate and sufficient, 18% (n = 9) accurate but insufficient and 2% were inaccurate and insufficient. In attempt 2, 86% (n = 43) of answers were accurate and sufficient, 12% (n = 6) were accurate but insufficient and 2% (n = 1) were inaccurate and insufficient. In attempt 3, 82% (n = 41) of answers were accurate and sufficient, 16% (n = 8) were accurate but insufficient and 2% (n = 1) were inaccurate and insufficient

From: Artificial intelligence derived large language model in decision-making process in uveitis