Longitudinal mass testing using rapid antigen detection tests (RADT) for serial screening of asymptomatic persons has been proposed for preventing SARS-CoV-2 community transmission. The feasibility of this strategy relies on accurate self-testing. To quantify the adequacy of serial self-performed SARS-CoV-2 RADT testing in the workplace, in terms of the frequency of correct execution of procedural steps and accurate interpretation of the range of possible RADT results. This prospective repeated cross-sectional study was performed from July to October 2021 at businesses with at least 2 active cases of SARS-CoV-2 infection in Montreal, Canada. Participants included untrained persons in their workplace, not meeting Public Health quarantine criteria (ie, required quarantine for 10 days after a moderate-risk contact with someone infected with SARS-CoV-2). Interpretation and performance were compared between participants who received instructions provided by the manufacturer vs those who received modified instructions that were informed by the most frequent or most critical errors we observed. Data were analyzed from October to November 2021. RADT testing using a modified quick reference guide compared with the original manufacturer's instructions. The main outcome was the difference in correctly interpreted RADT results. Secondary outcomes included difference in proportions of correctly performed procedural steps. Additional analyses, assessed among participants with 2 self-testing visits, compared the second self-test visit with the first self-test visit using the same measures. Overall, 1892 tests were performed among 647 participants, of whom 278 participants (median [IQR] age, 43 [31-55] years; 156 [56.1%] men) had at least 1 self-testing visit. For self-test visit 1, significantly better accuracy in test interpretation was observed among participants using the modified quick reference guide than those using the manufacturer's instructions for reading results that were weak positive (64 of 115 participants [55.6%] vs 20 of 163 participants [12.3%]; difference, 43.3 [95% CI, 33.0-53.8] percentage points), positive (103 of 115 participants [89.6%] vs 84 of 163 participants [51.5%]; difference, 38.1 [95% CI, 28.5-47.5] percentage points), strong positive (219 of 229 participants [95.6%] vs 274 of 326 participants [84.0%]; difference, 11.6 [95% CI, 6.8-16.3] percentage points), and invalid (200 of 229 participants [87.3%] vs 252 of 326 participants [77.3%]; difference, 10.0 [95% CI, 3.8-16.3] percentage points). Use of the modified guide was associated with improvements on self-test visit 2 for results that were weak positive (difference, 15.4 [95% CI, 0.7-30.1] percentage points), positive (difference, 19.0 [95% CI, 7.2-30.9] percentage points), and invalid (difference, 8.0 [95% CI, 0.8-15.4] percentage points). For procedural steps identified as critical for test validity, adherence to procedural testing steps did not differ meaningfully according to instructions provided or reader experience. In this cross-sectional study of self-performed SARS-CoV-2 RADT in an intended-use setting, a modified quick reference guide was associated with significantly improved accuracy in RADT interpretations.