Brown carbon (BrC) is a leading uncertainty in aerosol radiative forcing, and the source-resolved mass absorption coefficients (MAC) used in climate models come from both real-time (online) and filter-based (offline) measurements. Both are widely trusted, yet have rarely been applied to the same aerosol and compared source by source, leaving the assumption that they are interchangeable untested. Here we ask whether online and offline source apportionment yield the same sources of BrC absorption for wintertime urban aerosol. We combined a seven-wavelength aethalometer with high-resolution aerosol mass spectrometry (online, whole particle) and 24 h PM2.5 filter UV–Vis of water extracts (offline), attributing bulk absorption to positive-matrix-factorization factors by regression on each platform. The two platforms agreed on the bulk attribution (winter BrC carried by secondary and biomass-burning OA) and on consistency with worldwide observations, but inverted the source ranking: online, biomass burning and less oxidized secondary organic aerosol dominated while more oxidized OA (MO-OOA) was negligible (MAC 0.07 m² g⁻¹); offline, MO-OOA dominated (0.88 m² g⁻¹). This reversal reflects how the sample is measured, not an atmospheric change: water-soluble selection and a likely shift of MO-OOA toward reduced-nitrogen chromophores during offline preparation, while bulk composition and factor time series are preserved. The source inferred for BrC absorption is therefore strongly dependent on method and integration time, a critical caution for source-resolved BrC parameterizations that treat online and offline datasets as interchangeable.