AI-Generated Clinical Summaries

To the Editor In their recent Viewpoint, Dr Goodman and colleagues examined artificial intelligence (AI)–generated clinical summaries and identified concerns regarding variability, bias, and single-word errors in their AI summaries generated through the ChatGPT model. However, it must be noted that human writers exhibit these same shortcomings in clinical summarization. While striving for an “ideal” clinical summary that avoids variability, bias, and error is a worthwhile goal, requiring such idealized performance from AI is akin to discrimination. Instead, the quality of AI summaries should be compared against human performance rather than an idealization.