Background This meta‐narrative review, conducted according to the RAMESES (Realist And Meta‐narrative Evidence Syntheses: Evolving Standards) standards, critically examines the construct of self‐compassion to determine if it is an accurate target variable to mitigate work‐related stress and promote compassionate caregiving in healthcare providers. Methods PubMed, Medline, CINAHL, PsycINFO, and Web of Science databases were searched. Studies were coded as referring to: (1) conceptualisation of self‐compassion; (2) measures of self‐compassion; (3) self‐compassion and affect; and (4) self‐compassion interventions. A narrative approach was used to evaluate self‐compassion as a paradigm. Results Sixty‐nine studies were included. The construct of self‐compassion in healthcare has significant limitations. Self‐compassion has been related to the definition of compassion, but includes limited facets of compassion and adds elements of uncompassionate behavior. Empirical studies use the Self‐Compassion Scale, which is criticised for its psychometric and theoretical validity. Therapeutic interventions purported to cultivate self‐compassion may have a broader effect on general affective states. An alleged outcome of self‐compassion is compassionate care; however, we found no studies that included patient reports on this primary outcome. Conclusion We critically examine and delineate self‐compassion in healthcare providers as a composite of common facets of self‐care, healthy self‐attitude, and self‐awareness rather than a construct in and of itself.