This paper studies the relationship between patients' socio‐economic status and general practitioners' (GPs') service provision by exploiting administrative patient‐level data with information on consultation length, medical tests, and fee payments for each visit in Norway over a 5‐year period (2008–2012). To reduce patient heterogeneity, we limit the sample to a given condition, diabetes type II, that is treated almost exclusively in primary care. We estimate GP fixed‐effect models and control for a wide set of patient characteristics. Our results show that, for each visit, patients with low education get shorter consultations but more medical tests, patients with low income get less of both, and patients with low education/income get less services in monetary terms. We also find that, during a year, patients with low education/income visit the GP more often and receive more services in monetary terms. Thus, GPs treat patients differently according to their socio‐economic status, but we find no support for a social gradient.