AimsTo develop and preliminarily validate a measure of beliefs about compassion in health care and assess whether and which beliefs may predict compassion.DesignPre‐registered cross‐sectional online survey study with a repeated‐measures vignette component.MethodExploratory and Confirmatory Factor analyses were performed on a split sample of 890 healthcare professionals in Aotearoa New Zealand (NZ). Links with fears of compassion for others, burnout, trait compassion, compassion competency and ability and self‐efficacy were used to assess convergent and divergent validity. Linear mixed model regression analyses were used to assess relationships between beliefs and compassion. In writing this report, we adhered to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines.ResultsFour‐factor structure featuring three negative (compassion as harmful, not useful, draining) and one positive (compassion is important) type of beliefs was established. Confirmatory factor analysis indicated a good fit and subscales indicated good measures of validity. Internal consistency was achieved for the subset of beliefs (harmful, not useful). Regression analyses indicated negative effects of the belief that compassion is draining on caring, motivation to help and compassion overall; negative effects of the belief that compassion is not useful on the motivation to help and a positive effect of the belief that compassion is important on caring and compassion overall. There was no effect of beliefs that compassion is harmful on compassion measures.ConclusionThis report extends prior qualitative studies of beliefs about compassion in a large healthcare sample, offering a way to measure these potentially malleable factors that might be targeted in education, interventions and future research.Patient or Public ContributionThe study was designed in consultation with healthcare and compassion research professionals, including substantial input from Indigenous Māori healthcare professionals.