The El Niño–Southern Oscillation (ENSO) is the dominant mode of interannual climate variability and a cornerstone for climate model evaluation. Yet, the sensitivity of diagnosed ENSO teleconnections to the choice of observational datasets, reanalyses, and ENSO indices has received little attention. This study systematically evaluates boreal winter ENSO teleconnections using multiple atmospheric reanalyses, surface observational datasets, and three widely used ENSO indices: the Oceanic Niño Index, Multivariate ENSO Index, and Southern Oscillation Index. Statistical significance is assessed using Benjamini–Hochberg false discovery rate procedure to account for multiple spatially dependent hypothesis testing, providing a more rigorous assessment of teleconnection robustness than conventional approaches. Although the ENSO indices are highly correlated, the associated atmospheric and surface climate teleconnections differ substantially. ENSO teleconnections in upper-tropospheric (200 hPa) geopotential height are remarkably consistent across reanalyses. In contrast, teleconnections in mid-tropospheric (500 hPa) circulation, precipitation, and surface temperature are considerably more sensitive to the choice of reanalysis or observational dataset than to the choice of ENSO index. To examine the role of temporal variability, empirical orthogonal function and ensemble empirical mode decomposition analyses were applied to ERA5 geopotential height, CHIRPSv2 precipitation, and CRU land surface temperature to isolate variability at different temporal scales. For ERA5 upper-tropospheric geopotential height, ENSO teleconnections become markedly more robust after isolating the interannual component. Likewise, removing variability outside the interannual band generally strengthens teleconnections in precipitation and surface temperature, demonstrating that variability at other temporal scales can obscure ENSO-related signals. Conversely, restricting the analysis to the common temporal coverage shared by all datasets (1980–2018 for reanalyses and 1981–2022 for observational datasets) substantially reduces the statistical significance of ENSO teleconnections for nearly all variables, indicating that the overlapping observational record is too short for robust diagnosis. These results demonstrate that ENSO teleconnections are not uniquely defined climate features but depend strongly on dataset characteristics, like temporal scales represented and record length. Overall, the choice of observational or reanalysis dataset is more consequential than the choice of ENSO index. These findings underscore the need for longer, higher-quality observational records and caution against using surface ENSO teleconnections as unambiguous benchmarks for climate model validation.