This presentation is part of the KODAQS Data Quality Academy materials of the Competence Center for Data Quality in the Social Sciences (KODAQS). KODAQS is a joint initiative of GESIS – Leibniz Institute for the Social Sciences, the University of Mannheim, and the Ludwig Maximilian University of Munich (LMU) and is funded by the BMFTR and the EU NextGen Horizon Fund. The KODAQS Data Quality Academy program is designed to build practical and conceptual skills for assessing, documenting, and improving data quality across the entire lifecycle of quantitative research in the social sciences. The program combines core theoretical foundations with applied tools and domain-specific techniques. In this presentation, large language models (LLMs) are introduced as a tool in the social science research workflow. LLMs can be used for a multitude of possible tasks, but in order to use them effectively, an understanding of how they work is needed - both from a technical and data quality perspective. The presentation introduces the internal logic of LLMs and what they can be used for, as well as different types of LLMs, access options, and prompting strategies. Additionally, it discusses how these decisions impact data quality and explores mitigation strategies and reporting standards for better reproducibility. This presentation is the sixth part of a set of presentations on the topic of "Tools and Workflows".