Monthly potential evapotranspiration (PET) using the Thornthwaite (ThW) method fed with gridded climate datasets was evaluated relative to PET by ThW and to reference evapotranspiration (ETo) estimated using the FAO56 Penman-Monteith method (PM-FAO56) with observed data, for Rio de Janeiro state, and neighboring states in southeast Brazil. The PM-FAO56 method used monthly climate series (1961–2010) on sunshine hours, air temperature, relative humidity, and wind speed, from 21 weather stations of the National Institute of Meteorology (INMET), in Rio de Janeiro (RJ), São Paulo (SP), Minas Gerais (MG), and Espírito Santo (ES) to estimate ETo. The estimated PET using the ThW method was based on observed air temperature from INMET stations and also using gridded air temperature datasets (1961–2010) from the Global Historical Climatology Network (GHCN) and the University of Delaware (UDel). The ETP estimates had an agreement (Willmott index) between 0.54 ≤ dw ≤ 0.96 – GHCN and 0.62 ≤ dw ≤ 0.96 - UDel; a precision between 0.44 ≤ r² ≤ 0.88 – GHCN and 0.53 ≤ r² ≤ 0.95 - UDel; and RMSE inferior to 1.39 mm d− 1 – GHCN and 1.13 mm d− 1 - UDel. The PET estimated with the UDel gridded series was better overall performance than using the GHCN product. Differences in altitude, latitude and longitude were the main geographic variables determining the precision and agreement of the PET estimates using GHCN and Udel. The gridded datasets are an alternative for locations without climatic series data or with low-quality non-continuous data series.