This project studies whether changes in Swedish anti-discrimination legislation affect the content of job advertisements and, in turn, labour market outcomes. The empirical focus is on job advertisements from the Swedish Public Employment Service and their links to firms and workplaces over time. The project is conducted within Statistics Sweden’s secure research environment MONA, where linked administrative data on firms, workplaces, and employees are analysed. However, MONA does not provide GPU resources for large-scale inference with language models. The project requires access to GPU and therefore we will use Bianca/UPPMAX for the text-analysis parts, while the other analyses will be conducted in MONA. The project uses text analysis and machine-learning methods to classify the wording and content of a very large number of job postings. In particular, we study whether advertisements contain language that may be inclusive, exclusive, or potentially related to discrimination, for example with respect to age, immigrant background, gender, or disability. The classified advertisement data will then be linked back to firms and workplaces for further analysis within MONA. The broader goal is to understand whether stricter anti-discrimination rules influence how employers formulate job ads, whether firms change their recruitment behaviour, and whether these changes are associated with differences in who gets hired. This is important for assessing whether anti-discrimination policy affects not only formal legal compliance but also actual employer behaviour in the labour market. The project contributes to research on discrimination, recruitment, and labour market inequality by combining large-scale text data with administrative data.