The Scientific Process in the Age of Large Language Models

Scientific work is undergoing a fundamental transformation through large language models (LLMs) and multimodal AI systems. These technologies are evolving into comprehensive research assistants that accompany all phases of scientific activity — from the initial generation of ideas through theoretical orientation, methodological planning, and data processing to analysis, visualization, and the communication of results. As a result, the focus of scientific work is shifting away from mechanical routines toward more intellectual, theoretical, and interpretative activities. Already in the early phase of idea generation, LLMs expand the space of possible research questions by providing intellectual impulses, reorganizing concepts, and stimulating hypotheses. They help make potential research directions visible and create conceptual connections that facilitate entry into new fields of inquiry. The decision as to which of these approaches is ultimately viable and theoretically promising, however, remains with the scientist. In navigating the existing body of knowledge, AI systems are becoming tools that not only retrieve scientific literature but also structure and summarize it. They process large volumes of texts, condense their contents, and answer questions about them, thereby creating a clearer picture of the theoretical landscape. The ability to synthesize scientific work, extract key statements, and clarify follow-up questions immediately enables faster access to complex research fields. At the same time, it remains the researcher’s responsibility to critically assess the relevance and validity of this information.