Abstract: The boom in ESG investing has created a demand for information. ESG raters have stepped up to fill this gap, but are under scrutiny for inaccurate ratings. This inaccuracy is often measured by the divergence (low correlation) of ratings. We provide a model of ESG ratings competition, where raters provide information about multiple (unrelated) subcategories and set fees. The social optimum may be for the raters to specialize in different subcategories, while the competitive outcome may be for them to generalize among subcategories. Generalizing increases their stand-alone value and, hence, their pricing power. We also demonstrate that specialization maximizes divergence among ratings. Hence, divergence may be a poor measure of welfare. Furthermore, ratings inaccuracy may lead to the underprovision of effort by firms to improve their ESG performance. The results are robust to heterogeneity in ESG preferences, changing the weight investors give to ratings, and varying the information structure of the ratings.