Many diabetic patients remain undiagnosed, and oral findings may offer an unrealized opportunity for the identification of affected individuals unaware of their condition. We recruited 601 individuals who presented for care at a dental clinic, were ≥40 years old, if non-Hispanic white, and ≥30 years old, if Hispanic or non-white, and had never been told they have pre-diabetes or diabetes. Those with at least one self-reported diabetes risk factor (N=535) received a periodontal examination and a point-of-care hemoglobin A1c (HbA1c) test. A fasting plasma glucose (FPG) test was used as the study outcome, signifying potential diabetes or pre-diabetes. Performance characteristics of simple models of dysglycemia (FPG≥100 mg/dL) identification were evaluated and optimal cut-offs identified. A model including only two dental variables had an estimated area under the receiver operating characteristic curve (AUC) of 0.65. The addition of a point-of-care HbA1c test improved the AUC to 0.79 (p<0.001). The presence of ≥26% deep pockets or ≥4 missing teeth correctly identified 73% of true cases; the addition of an HbA1c≥5.7% increased correct identification to 92%. Analysis of our data suggests that oral healthcare professionals have the opportunity to identify unrecognized diabetes and pre-diabetes in dental patients and refer them to a physician for further evaluation and care.