Large language models can identify judgmental language in clinical notes, but the settings play a major role in accuracy.