BackgroundHIV status disclosure is an important decision with barriers specific to young men who have sex with men (YMSM), who have the highest rates of new HIV infections in the United States. Behavioral and social determinants of the difficulty to disclose can include fear of rejection, stigma, loss of financial stability, and lack of communication skills. Once able to disclose, a person may have increased access to social support and improved informed risk reduction conversations and medication adherence. Despite the known challenges and advantages of disclosure, there are few effective tools supporting this behavior.ObjectiveTo address this gap in disclosure interventions, the Tough Talks (TT) app, an mHealth intervention using artificial intelligence (AI)–facilitated role-playing scenarios, was developed for YMSM. This paper reports stages of development of the integrated app and results of the usability testing.MethodsBuilding on the successful development and testing of a stand-alone interactive dialogue feature in phases 1-3, we conducted additional formative research to further refine and enhance the disclosure scenarios and develop and situate them within the context of a comprehensive intervention app to support disclosure. We assessed the new iteration for acceptability and relevance in a usability study with 8 YMSM with HIV. Participants completed a presurvey, app modules, and a semistructured qualitative interview.ResultsTT content and activities were based on social cognitive theory and disclosure process model framework and expanded to a 4-module curriculum. The AI-facilitated scenarios used dialogue from an utterance database developed using language crowdsourced through a comic book contest. In usability testing, YMSM reported high satisfaction with TT, with 98% (31/33) of activities receiving positive ratings. Participants found the AI-facilitated scenarios and activities to be representative and relevant to their lived experiences, although they noted difficulty having nuanced disclosure conversations with the AI.ConclusionsTT was an engaging and practical intervention for self-disclosure among YMSM with HIV. Facilitating informed disclosure decisions has the potential to impact engagement in sexual risk behaviors and HIV care. More information is needed about the ideal environment, technical assistance, and clinical support for an mHealth disclosure intervention. TT is being tested as a scalable intervention in a multisite randomized controlled trial to address outstanding questions on accessibility and effect on viral suppression.Trial RegistrationClinicalTrials.gov NCT03414372; https://clinicaltrials.gov/ct2/show/NCT03414372