PURPOSE To evaluate the inter- and intraobserver reliability of the CADS score, a previously described facial nerve grading instrument for ophthalmic grading of facial nerve palsy. METHODS This is a cross-sectional validation study. Two clinicians independently assessed and graded each patient on the same day, masked to each other's grading. Four parameters are assessed in the CADS scale: Cornea (0-3, ±a), static Asymmetry (0-2), Dynamic function (0-3), and Synkinesis (0-2). One clinician reassessed the patients and performed the grading again at a minimum time interval of 1 hour later. A weighted κ analysis was performed to determine inter- and intraobserver reliability using 95% bootstrapped bias-corrected and accelerated (BCa) confidence intervals (CIs). RESULTS Thirty-three patients (27 women, mean age 51.7, range 23-80 years) with unilateral facial nerve palsy were graded. The overall interobserver reliability was 0.80 (95% BCa CI: 0.68-0.91) for cornea, 0.93 for resting asymmetry (95% BCa CI: 0.55-1.00), 0.80 for dynamic function (95% BCa CI: 0.50-0.96), and 0.88 (95% BCa CI: 0.71-0.96) for synkinesis. The overall intraobserver reliability was 0.93 for cornea (95% BCa CI: 0.83-0.98), 0.82 for resting asymmetry (95% BCa CI: 0.53-0.96), 0.92 for dynamic function (95% BCa CI: 0.72-1.00), and 0.98 for synkinesis (95% BCa CI: 0.84-1.00). CONCLUSION The CADS grading scale demonstrates good interobserver reliability and very good intraobserver reliability. It incorporates all ophthalmic complications for facial nerve palsy and remains easy to use and refer to.