Objective: Pain is common in the elderly. Knee pain may predict knee cartilage loss, but whether generalized pain is associated with knee cartilage loss is unclear. This study, therefore, aimed to determine whether pain at multiple sites predicts knee cartilage volume loss among community-dwelling older adults, and, if so, to explore potential mechanisms. Methods: Data from the prospective Tasmanian Older Adult Cohort study was utilized (n = 394, mean age 63 years, range 52–79 years). Experience of pain at multiple sites was assessed using a questionnaire at baseline. T1-weighted fat-saturated magnetic resonance imaging of the right knee was performed to assess the cartilage volume at baseline and after 2.6 years. Linear regression modeling was used with adjustment for potential confounders. Results: The median number of painful sites was 3 (range 0–7). There was a dose-response relationship between the number of painful sites and knee cartilage volume loss in the lateral and total tibiofemoral compartments (lateral β = −0.28% per annum; total β = −0.25% per annum, both P for trend < 0.05), but not in the medial compartment. These associations were stronger in participants without radiographic knee osteoarthritis (OA) (P < 0.05) and independent of age, sex, body mass index, physical activity, pain medication, and knee structural abnormalities. Conclusion: The number of painful sites independently predicts knee cartilage volume loss, especially in people without knee OA, suggesting that widespread pain may be an early marker of more rapid knee cartilage loss in those without radiographic knee OA. The underlying mechanism is unclear, but it is independent of anthropometrics, physical activity, and knee structural abnormalities.