<p>The vibrational frequency of carbon monoxide (CO) adsorbed on ceria-based catalysts serves as a sensitive probe for identifying exposed surface facets, provided that experimental reference data on well-defined single-crystal surfaces and reliable theoretical assignments are available. Previous studies have shown that the hybrid DFT approach using the HSE06 functional yields good<br>agreement with experimental observations, whereas the generalized gradient approximation (GGA) with PBE+U does not. In this work, we assess the performance of different exchange-correlation functionals by comparing the meta-GGA functionals SCAN and r2SCAN meta-GGA functionals with HSE06 in predicting CO vibrational frequencies on cerium oxide surfaces. The meta-GGA<br>functionals offer no significant improvement for oxidized CeO2(111) and CeO2(110) surfaces and fail to localize excess charge on the reduced surfaces. Adding a Hubbard U term improves charge localization, but the predicted vibrational frequencies still fall short of HSE06 accuracy. These limitations are attributed to the meta-GGA's inability to adequately capture facet- and configuration-specific donation and back-donation effects, which influence the C—O bond length and CO force constant upon adsorption. Despite the higher computational cost when used with plane-wave basis sets, hybrid DFT remains essential for accurate interpretation of experimental results.</p>