Prompt diagnosis of group A streptococcal pharyngitis is essential for primary prevention of acute rheumatic fever and rheumatic heart disease, yet affordable point-of-care diagnostics remain limited in low-resource settings. We conducted a prospective study among Ugandan children 5-15 years of age presenting with sore throat, comparing smartphone-acquired oropharyngeal videos analyzed using an artificial intelligence-based image recognition model-trained and validated using combined data from Ugandan and US-based cohorts-with microbiologic culture as the reference standard. Among 82 children evaluated with the optimized model, 12 (15%) had culture-confirmed group A streptococcal pharyngitis. The smartphone-based approach demonstrated 100% sensitivity (95% confidence interval [CI]: 73.5%-100%), 95.7% specificity (95% CI: 88.0%-99.1%) and an overall accuracy of 96.6% (95% CI: 90.0%-99.3%), with an area under the receiver operating curve of 0.93 (95% CI 0.83-1.00). These findings support the potential of smartphone-based artificial intelligence as a rapid, noninvasive diagnostic tool for primary prevention in rheumatic heart disease-endemic settings.