Background: Social prescribing (SP) link workers (LWs) listen to patients’ concerns and difficulties, and connect them to relevant community assets (groups, organisations, or charities) that can help with their non-medical issues (for example, loneliness, debt, housing). LW retention is key to sustaining SP within primary care. Aim: To examine occupational self-efficacy, job discrepancy, and other factors as potential predictors of LWs’ intentions to leave or remain in their posts. Design & setting: Cross-sectional survey involving LWs from the United Kingdom. Method: An online questionnaire was distributed via SP-related organisations. Questions were on the following: (a) intention to leave the role; (b) demographics; and (c) role experience, including occupational self-efficacy and discrepancy between expectations and reality of the job. Questions were mainly closed, although some allowed LWs to provide a written response. Logistic regression models were fitted to identify predictors, and content analysis used to categorise open-ended responses. Results: In total, 342 questionnaire responses were included in the analysis. Higher job discrepancy was associated with past (odds ratio [OR] per 30 unit increase = 6.86; 95% confidence interval [CI] = 3.91 to 12.0; P<0.001) and future (OR = 4.86; 95% CI = 2.70 to 8.72; P<0.001) intentions to leave, while lower occupational self-efficacy was associated only with past intentions to leave (OR per 10 unit decrease = 1.91; 95% CI = 1.24 to 2.93; P = 0.003). Conclusion: Findings highlight factors influencing LW retention, offering a foundation for targeted interventions, which could include clearer communication about the role during recruitment, and adjusting job descriptions and support when required.