The study examined the factor structure of burnout, as measured with the Oldenburg Burnout Inventory. The participants were 235 employees of a public administration agency who assessed their burnout online for 10 consecutive working days. Two models were tested with multilevel confirmatory factor analysis, assuming the same one or two-factor structure at the within- and between-person levels. Both models showed a reasonable fit to the data, but due to a strong correlation between exhaustion and disengagement and low within-person reliability for disengagement, a unidimensional model seems more valid. A cross-level invariance was not confirmed for either of the structures, showing that factor loadings for the same items differ significantly between the levels. This suggests that burnout is not the same latent variable at each level; rather, there are factors other than daily burnout that influence person-level scores and ignoring these across-level discrepancies may lead to biased conclusions.