Citation screening sorts research-paper titles and summaries into potentially relevant, irrelevant or uncertain. Noel-Storr and colleagues’ report (26 April 2021) compared trained volunteers with prior screening experience against a review team on 9,546 records. The crowd finished 33 hours after launch; the team’s reported completion time was 410 hours. Elapsed time is time until completion; total labor adds everyone’s working hours. The reported team total does not clearly specify which time basis it uses. A classification is one person’s judgment; records received several. Reported per-judgment averages—logged judgment time divided by judgments—were roughly six seconds for the crowd versus four for the team. Training design, building and piloting were omitted from the speed comparison. This single maternity-care task establishes neither total labor savings nor capacity to scrutinize unfamiliar AI-generated research.