There is an increasing need for analysing detailed companies’ demand for workers, that is for occupation, skills or competences and qualifications. Current surveys conducted by Statistical Offices do not contain information on demand of companies for future workers’ competences or qualifications. One might consider online job offers as support or alternative for surveys on vacancy market. However, these data sources are unstructured and relevant information should be extracted to conduct quality and representativeness assessment and estimation process. Our main contribution is the proposal of an efficient method for analysing Internet job offers in the context of detailed information they contain. The method is based on gathering Internet job offers and analysing them with text mining and machine learning tools. We present caveats behind such a research and propose solutions to them. We apply this method for the Polish vacancy market and compare results with ongoing representative surveys to correct non-probability character of these data. The results are especially important for economists, education sector, and labour market institutions, e.g. for shaping the OECD Skills Strategy. Such detailed information may be used to adjust labour market and education policy, especially directed to reducing the structural unemployment.