Providing eyewitness testimony involves monitoring one's memory to provide a detailed and accurate account: reporting details likely to be accurate and withholding potentially inaccurate details. Autistic individuals reportedly experience difficulties in both retrieving episodic memories and monitoring their accuracy, which has important implications for eyewitness testimony. Thirty autistic and 33 IQ‐matched typically developing (TD) participants viewed a video of a mock bank robbery followed by three phases of questions (with judgments of confidence). In Phase 1, participants freely generated the granularity of their responses (i.e., fine‐ or coarse‐grained). In Phase 2, participants answered the same questions but provided both a fine‐ and a coarse‐grained answer. In Phase 3, participants were instructed to maximize accuracy over informativeness by selecting one of their Phase 2 answers as their final answer. They either received the questions socially (from the experimenter) or answered them online. There were no group differences in accuracy or metacognitive monitoring, with both autistic and TD witnesses demonstrating: (a) a strong preference for reporting fine‐grained details at the expense of accuracy; (b) improved though still suboptimal grain size reporting when instructed to maximize accuracy over informativeness; (c) effective accuracy monitoring; and (d) higher overall accuracy when questions were delivered socially. There was, however, a subtle difference in metacognitive control, with autistic witnesses performing more poorly than TD witnesses when questions were delivered socially, but not when they were delivered online. These findings contrast with evidence suggesting that autism is marked by impairments in episodic memory and metacognitive monitoring and control.