Background Clinical prediction rules can assist clinicians to identify patients with low back pain (LBP) who are likely to recover quickly with minimal treatment; however, there is a paucity of validated instruments to assist with this task. Method We performed a pre-planned external validation study to assess the generalizability of a simple 3-item clinical prediction rule developed to estimate the probability of recovery from acute LBP at certain time points. The accuracy of the rule (calibration and discrimination) was determined in a sample of 956 participants enrolled in a randomized controlled trial. Results The calibration of the rule was reasonable in the new sample with predictions of recovery typically within 5-10% of observed recovery. Discriminative performance of the rule was poor to moderate and similar to that found in the development sample. Conclusions The results suggest that the rule can be used to provide accurate information about expected recovery from acute LBP, within the first few weeks of patients presenting to primary care. Impact analysis to determine if the rule influences clinical behaviours and patient outcomes is required.