Objective To investigate the association between atlanto-occipital radiographic alignment in flexion and cervical spondylosis (CS). Methods This is a retrospective case-control study. CS patients were recruited from our hospital, and the age/gender/body mass index (BMI)-matched healthy controls were selected from the subjects in health examinations at the same hospital between January 2015 and May 2019. A total of 464 subjects was included in the study. There are 282 males and 182 females. The ages of patients were 20 to 67 years, and the mean age was 33.9 years. CS patients were considered the case group. Based on surgical treatments, they were subdivided into non-operation group and operation group. The operation group and non-operation group had 45 and 187 patients, respectively, while 232 subjects were included in the control group. The angle between McGregor's line and C-1 line (O-C-1 angle) was evaluated on images taken in flexion (F-OC) and neutral positions (N-OC) independently. The relationship between the FOC (FOC=F-OC-N-OC) and Neck Disability Index (NDI) was examined, and the involvement of the FOC in the onset of CS was analyzed. Receiver operating characteristic (ROC) curve analysis was performed to determine the optimal cut-off for detecting an increased risk of CS. Results The median follow-up time was 51.6 months (25-115 months). The case groups, especially the operation group, tended to be older (55.8 +/- 11.2 vs 41.6 +/- 13.8 vs 23.5 +/- 5.5 years, P < 0.001), have a higher NDI score (12.2 +/- 4.5 vs 6.2 +/- 2.1 vs 3.2 +/- 1.2, P < 0.001), and longer medical history (10.5 +/- 9.5 vs 6.8 +/- 11.2 years, P < 0.001). One-way analysis of variance showed statistically significant differences in FOC between the control and case groups (1.4 degrees +/- 1.2 degrees vs 3.6 degrees +/- 1.9 degrees vs 7.2 degrees +/- 2.0 degrees, P < 0.001). Besides, a post-hoc Tukey test showed a lower FOC in the operation group compared with that in the non-operation group (1.4 degrees +/- 1.2 degrees vs 3.6 degrees +/- 1.9 degrees, P < 0.001). Using FOC as a radiological predictive model to predict CS, the cut-off value was 4.2 degrees. Using FOC as a radiological predictive model to predict CS, the area under the curve (AUC) was 0.86 (95% CI: 0.78-0.92, P < 0.001). In the univariable risk analysis model, conditional logistic regression showed that the FOC level was an independent factor with an important role in the risk of CS. The odds rose to 8.2 times when FOC reached the level under 4.2 degrees (OR = 8.2; 95% CI: 6.4-10.0; P < 0.001). There existed a significant negative correlation between FOC levels and NDI (r = -0.451, P = 0.016). Conclusions Stiff O-C-1, which is defined as FOC <= 4.2 degrees, represented decreased flexion dysfunction of atlanto-occipital joint and is closely associated with high risk for the occurrence of CS. This finding could show a possible relationship between upper and lower cervical spine and help spine surgeons to understand the pathological process of CS and implement appropriate management.