Abstract There are notable discrepancies between near‐surface humidity trends in observational products as compared to climate models. Here, we explore the extent to which the discrepancy can be explained by inhomogeneities in the observational record. To do so, we apply a validated changepoint detection algorithm to the residual between two observational products, the ERA5 reanalysis and HadISDH station‐based product, and a climate model simulation whose circulation is constrained to match the observations. Based on synthetic data tests, the approach is likely to identify changepoints in excess of 0.32 g/kg, and rarely identifies a changepoint when none is present. For ERA5, we identify changepoints across 31% of the global land area; qualitatively similar results are found for HadISDH. Correcting the inhomogeneities reduces, but does not eliminate, the previously identified discrepancy between model simulations and the historical record.