The vagally-mediated heart rate variability (HRV), an indicator of the autonomic nervous system (ANS), has been proposed as a transdiagnostic marker for emotion regulation (ER). In people with psychotic disorders (PSY), HRV is profoundly reduced compared to healthy controls (HC). Similarly, questionnaire-based assessments of adaptive ER point to a deficit in PSY. To address HRV as a potential marker for ER in psychosis, we investigated a large sample including PSY (n = 40) and HC (n = 32) as well as vulnerable (n = 19 clinical high-risk) and clinical (n = 28 anxiety disorders) controls. We tested the differential effectiveness of an instructed adaptive ER strategy, whether resting-state HRV predicts adaptive ER, and whether HRV serves as a state index of ER effort. The participants repeatedly played a social exclusion ballgame while they applied either cognitive reappraisal or no regulation ("just-play") in randomized order. PSY displayed overall higher levels of negative affect and paranoia than HC but both groups applied reappraisal successfully (i.e., more benefit: lower negative affect and paranoia after reappraisal compared to "just-play"). Resting-state HRV did not predict successful reappraisal in the total sample. However, within PSY, a higher resting-state HRV predicted more benefit from reappraisal. State HRV did not differ between the reappraisal and "just-play" condition. Contrary to our expectations, participants with psychosis applied an instructed adaptive ER strategy successfully. As expected, the ANS marker of HRV predicted that benefit; however, this was only the case in PSY. Overall, HRV was not a robust but a tentative marker of ER in the present investigation.