Objective To determine the diagnostic accuracy of the Addenbrooke’s Cognitive Examination (ACE-R) for older adults with low education, without dementia, in two capitals in northeastern Brazil, compared to subjects with MCI. Methods 100 participants were collected from a previous neurological and psychiatric evaluation and were subsequently subjected to the ACE-R. Among them, 18 subjects with amnestic mild cognitive impairment (aMCI), 22 with non-amnestic mild cognitive impairment (naMCI), and 60 healthy controls. Results Optimal ACE-R accuracy yielded excellent values for the comparison between controls and naMCI [Area Under the Curve (AUC) = 0.919)] and controls and aMCI (AUC= 0.921); conversely, very fair accuracy was reported for the comparison between aMCI and naMCI (AUC= 0.578). Conclusions These findings support establishing reliable cutoff scores for cognitive assessment of elderlies with low schooling and cognitive decline, not dementia, taking into consideration ecological and regional variables.