Objective: To describe current practice and investigate factors associated with selection for rehabilitation following acute stroke. Design: Prospective observational cohort study. Setting: Seven public hospitals in Queensland, Australia. Subjects: Consecutive patients surviving acute stroke. Measures: Rehabilitation selection processes are assessment for rehabilitation needs, referral for rehabilitation and receipt of rehabilitation. Functional impairment following stroke is modified Rankin Scale (mRS). Results: We recruited 504 patients, median age 73 years (interquartile range (IQR) = 62–82), between July 2016 and January 2017. Of these, 90% (454/504) were assessed for rehabilitation needs, 76% (381/504) referred for rehabilitation, and 72% (363/504) received any rehabilitation. There was significant variation in all rehabilitation selection processes across sites (P < 0.05). In multivariable analyses, stroke unit care (odds ratio (OR) = 2.7; 95% confidence interval (CI) = 1.1, 6.6) and post stroke functional impairment (severe stroke mRS 4–5: OR = 10.9; 95% CI = 4.9, 24.6) were associated with receiving an assessment for rehabilitation. Receipt of rehabilitation was more likely following assessment (OR = 6.5; 95% CI = 2.9, 14.6) but less likely in patients with dementia (OR = 0.2; 95% CI = 0.1, 0.9), end-stage medical conditions (OR = 0.4; 95% CI = 0.2, 0.8) or ischaemic stroke (OR = 0.4; 95% CI = 0.1, 0.9). The odds of receiving rehabilitation increased with greater impairment: OR = 3.0 (95% CI = 1.5, 4.9) for mRS 2–3 and OR = 12.5 (95% CI = 6.5, 24.3) for mRS 4–5. Among patients with mild-moderate impairment (mRS 2–3), 39/117 (33%) received no rehabilitation. Conclusions: There was significant inter-site variation in rehabilitation selection processes. The major factors influencing rehabilitation access were assessment for rehabilitation needs, co-morbidities and post-stroke functional impairment. Gaps in access to rehabilitation were found in those with mild to moderate functional impairment.