This study investigates the determinants and dynamics of the linguistic features of key audit matters (KAMs) in European countries. Using natural language processing algorithms, including FINBERT, I quantify stylistic and content-based text characteristics at the KAM level and find that KAM length, readability, sentiment, quantitative density, specificity, the degree of forward-looking statements, and the extent of boilerplate language are associated with the type of KAM topics, client attributes, and audit firm characteristics. In additional analyses, I also find early empirical evidence of a time trend in these linguistic features. Since their introduction in 2016, KAMs are becoming longer, more quantitative, more specific, but also include more boilerplate phrases. Collectively, the results of the study contribute to a more nuanced understanding of the determinants and dynamics of KAM disclosures.
