Abstract Objectives Loneliness in people who experience psychosis is common and associated with poor mental health. In this randomised trial, we tested the feasibility and acceptability of an adapted Groups for Health (G4H) intervention for loneliness, delivered in group or individual format. Design Mixed methods, two‐arm feasibility randomised controlled trial. Methods Forty individuals who self‐identified as having psychosis were recruited from UK mental health care services, recovery colleges and charities. G4H was modified for people with psychosis, with participants randomised to receive the intervention delivered via group ( N = 20) or individual ( N = 20) format. The primary outcomes related to trial acceptability and feasibility. Exploratory repeated measures ANOVAs and t‐tests evaluated differences between formats over time in loneliness, wellbeing and possible mechanisms of change including social identification, identity integration and perceived in‐group and out‐group empathy. Measures were completed at baseline, end of treatment and 1‐ and 6‐month follow‐up. Results Recruitment, retention and trial acceptability ratings for both group and individual formats of G4H were acceptable to good. No participants reported experiencing a serious adverse event. Exploratory ANOVAs indicated no differences related to format but positive change in key variables of loneliness, wellbeing, social identification and identity integration over time. T ‐tests for loneliness indicated that this change was step‐wise from baseline, through end of treatment to 1‐month follow‐up. Conclusions G4H is a feasible intervention for people with psychosis who identify as lonely and it can be delivered in either group or individual formats. This feasibility trial provides support for a future full randomised controlled trial.