The eligibility criteria currently used for neurodevelopmental follow-up are traditionally based on perinatal events and characteristics of the infants at birth. However, they seem unsatisfactory to target efficiently all children who will manifest long-term neurologic sequelae and eventually require rehabilitation services. The updated version of the Amiel-Tison's Neurological Assessment At Term (ATNAAT) is expected to allow a better prediction of the neurodevelopmental outcome in high-risk infants. The main objective of the present study, which was performed on 35 infants, was to analyze the interobserver reliability of the updated version of ATNAAT. The evaluator and the observer coded the items of the test simultaneously. Among the 35 items tested, 16 demonstrated an excellent reliability based on the kappa coefficient, 11 items yielded a fair to good reliability, whereas only two items produced an agreement below 0.40. The final synthesis, which was a global appreciation of the neurologic status based on the different findings, yielded a good reliability with a kappa coefficient of 0.76. Among the infants who had a nonoptimal outcome from the assessment, only 38.5% met the traditional criteria currently used for follow-up. It would be important to conduct further research on predictive validity to demonstrate the capacity of the ATNAAT to forecast the long-term neurologic outcome of infants at risk.