We analyzed inter-individual variability in response to exercise among acutely hospitalized oldest-old adults. In this ancillary analysis of a randomized controlled trial, 268 patients (mean age 88 years) were assigned to a control (n = 125, usual care) or intervention group (n = 143, supervised exercise, i.e., walking and rising from a chair [1–3 sessions/day]). Intervention group patients were categorized as responders, non-responders, or adverse responders (improved, no change, or impaired function in activities of daily living [ADL, Katz index] from hospital admission to discharge, respectively). We analyzed the association between responsiveness to exercise and variables assessed at baseline (2 weeks pre-admission), admission, during hospitalization, at discharge, and during a subsequent 3-month follow-up. An impaired ADL function and worse nutritional status at admission were associated to a greater responsiveness, whereas a better ADL function at admission, longer hospitalization and lower comorbidity index were associated with a poorer response (p < 0.05). Adverse responders had worse outcomes at discharge and during the follow-up (e.g., impaired physical performance and greater fall number) (p < 0.05). Although exercise intervention helps to prevent ADL function decline in hospitalized oldest-old people, a number of them—particularly those with a better functional/health status at admission and longer hospitalization—are at higher risk of being adverse responders, which can have negative short/middle-term consequences.