Importance: The number of olfactory dysfunction cases has increased dramatically because of the COVID-19 pandemic. Identifying therapies that aid and accelerate recovery is essential. Objective: To determine the efficacy of bimodal visual-olfactory training and patient-preferred scents vs unimodal olfactory training and physician-assigned scents in COVID-19 olfactory loss. Design, Setting, and Participants: This was a randomized, single-blinded trial with a 2-by-2 factorial design (bimodal, patient preferred; unimodal, physician assigned; bimodal, physician assigned; unimodal, patient preferred) and an independent control group. Enrollment occurred from February 1 to May 27, 2021. Participants were adults 18 to 71 years old with current olfactory loss defined as University of Pennsylvania Smell Identification Test (UPSIT) score less than 34 for men and less than 35 for women and duration of 3 months or longer. Olfactory loss was initially diagnosed within 2 weeks of COVID-19 infection. Interventions: Participants sniffed 4 essential oils for 15 seconds with a 30-second rest in between odors for 3 months. Participants in the physician-assigned odor arms trained with rose, lemon, eucalyptus, and clove. Participants randomized to the patient-preferred arms chose 4 of 24 available scents. If assigned to the bimodal arm, participants were shown digital images of the essential oil they were smelling. Main Outcomes and Measures: The primary end point was postintervention change in UPSIT score from baseline; measures used were the UPSIT (validated, objective psychometric test of olfaction), Clinical Global Impressions Impression-Improvement (CGI-I; self-report improvement scale), and Olfactory Dysfunction Outcomes Rating (ODOR; olfaction-related quality-of-life questionnaire). Results: Among the 275 enrolled participants, the mean (SD) age was 41 (12) years, and 236 (86%) were female. The change in UPSIT scores preintervention to postintervention was similar between the study arms. The marginal mean difference for change in UPSIT scores preintervention to postintervention between participants randomized to patient-preferred vs physician-assigned olfactory training was 0.73 (95% CI, -1.10 to 2.56), and between participants randomized to bimodal vs unimodal olfactory training was 1.10 (95% CI, -2.92 to 0.74). Five (24%) participants in the control arm had clinically important improvement on UPSIT compared with 18 (53%) in the bimodal, patient-preferred arm for a difference of 29% (95% CI, 4%-54%). Four (19%) participants in the control group self-reported improvement on CGI-I compared with 12 (35%) in the bimodal, patient-preferred arm for a difference of 16% (95% CI, -7% to 39%). The mean change in ODOR score preintervention to postintervention was 11.6 points (95% CI, 9.2-13.9), which was not deemed clinically important nor significantly different between arms. Conclusions and Relevance: Based on the change in UPSIT scores, this randomized clinical trial did not show any difference between intervention arms, but when exploring within-patient change in UPSIT as well as self-reported impression of improvement, active interventions were associated with larger improvement than controls with a potential advantage of bimodal intervention. While not definitive, these results suggest that patients with COVID-19 olfactory loss may benefit from bimodal visual-olfactory training with patient-preferred scents.