Abstract
This study evaluates 87 established AI-diagnostic mHealth apps intended for primary healthcare advice without direct physician consultation. We developed a conceptual framework examining 40 app metrics and features across six constructs: social proof, capabilities, ease-of-use, credibility, privacy/security, and monetization. Hierarchical regression analysis tested whether these attributes explain market adoption (cumulative downloads), supplemented by sentiment and content analyses of 166,303 user reviews. The dataset shows a dominance of dermatology and symptom checker apps. Regression results indicate that downloads are positively associated with user star evaluations (social proof), professional endorsements, greater app age, more extensive app-store descriptions (credibility), and in-app purchases (monetization). Sentiment analysis revealed 81.2% positive, 9.7% negative, and 9.2% neutral user reviews. Content analysis of reviews matched to the constructs showed that users value non-judgmental real-time feedback. However, users also highlighted concerns such as pricing opacity, technical glitches, and third-party data sharing. To ensure methodological rigor, the regression results were assessed for robustness, while both the regression and content analysis findings were face validated with app developers, physicians, health policymakers, and technology attorneys. Although the qualitative and quantitative findings do not fully converge, they jointly provide actionable insights for technology developers and healthcare organizations to improve the design of AI-based diagnostic applications and develop robust evaluation criteria.