I conduct the first systematic comparison of ESG ratings between China’s two leading providers, CSMAR and CNRDS, using a matched panel of A-share firms from 2015 to 2020. I perform a wide-ranging empirical analysis at firm, industry, and temporal levels to investigate whether the two systems capture the same latent construct associated with corporate sustainability. My findings highlight significant and persistent divergence between rating levels, correlation structures, and reliability measures. CSMAR consistently assigns significantly higher ESG scores than CNRDS, particularly in the Environmental and Social dimensions, and cross-system correlations are extremely weak and frequently negative in the Social pillar. Measures of agreement, including correlation coefficients and concordance metrics, reveal sharp divergences in firm-level rankings between the two systems. Critically, I demonstrate that standard normalization methods, including percentile ranking, min–max scaling, and industry-adjusted standardization, fail to reconcile these differences. Even after removing scale and industry effects, rank-order disagreement remains pronounced, suggesting that the divergence reflects fundamental methodological differences rather than distributional or scaling artifacts. It is especially pronounced in industries with complex, qualitative, and disclosure-intensive ESG profiles, such as Finance and Information Technology, where greater measurement discretion amplifies methodological divergence. Collectively, the evidence indicates that CSMAR and CNRDS are non-interchangeable proxies for distinct ESG constructs. These findings have important implications for empirical research design, ESG-based investment strategies, and corporate sustainability assessment in China, highlighting the necessity of treating ESG rating choice as a core methodological decision rather than a neutral data input