Proofreading is an important cognitive skill, yet methods for enhancing error detection have received little research attention. We report two experiments comparing the efficacy of proofreading aloud (i.e., "production")-a common proofreading tip-versus in a disfluent font (Sans Forgetica) purported to introduce a desirable difficulty. Participants read eight short texts silently, eight aloud, and eight in the disfluent font-and recorded the noncontextual errors (i.e., typos) and contextual errors (i.e., grammar, word choice) that they detected. Relative to silent reading, proofreading aloud improved detection of both types of errors, whereas the disfluent font impaired detection of noncontextual errors and did not aid detection of contextual errors. Metacognitively, participants did not expect the benefit of reading aloud, whereas they expected the disfluent font to impair proofreading. Proofreading aloud works, and invites further research on how it works, whereas there was no indication that proofreading in a disfluent font is useful.