BACKGROUND: Substance use is associated with higher rates of antiretroviral non-adherence and poor HIV outcomes. This study examined how HIV care providers assess substance use, and which questions elicit accurate patient disclosures. METHODS: We conducted a conversation analysis of audio-recorded encounters between 56 providers and 162 patients living with HIV (PLWH) reporting active substance use in post-encounter interviews (cocaine or heroin use in the past 30 days, > 4 days intoxicated in past 30 days, or AUDIT score ≥ 8). We assessed the frequency of substance use discussion, characterized the types of questions used by providers, and determined the frequency of accurate patient disclosure by question type. RESULTS: In 55 reports of active substance use, providers already knew about the use (n = 16) or patients disclosed unpromptednn = 39). Among the remaining 155 instances of substance use in which providers had the opportunity to elicit disclosure, 78 reports (50 %) of substance use were not discussed. Of the remaining 77 reports in which the provider asked about substance use, 55 (71 %) patients disclosed and 22 (29 %) did not disclose. Questions were classified as: open-ended (n = 18, BHow's the drinking going?^); normalizing (n = 14, BWhen was the last time you used?^); closed-ended (n = 36, BHave you used any cocaine?^); leading towards non-use (n = 9, BHave you been clean?^). Accurate disclosure followed 100 % of open-ended and normalizing questions, 58 % of closed-ended questions, and 22 % of leading questions. After adjusting for drug type, closed-ended questions were 41 % less likely (p < 0.001), and 'leading' questions 78 % less likely (p = 0.016) than broad and normalizing questions to elicit disclosures. CONCLUSION: Providers in this sample missed almost half of the opportunities to identify and discuss substance use with PLWH. Providers can increase the probability of patient disclosure by using open-ended or normalizing questions that ask about the Blast time^that the patient used drugs or alcohol.