<p>Large Language Models (LLMs) are increasingly used across many scientific domains, but their role in fisheries management research remains poorly defined. Here we provide an evidence-based overview of how LLMs might be used, and what risks they introduce, in fisheries management research. First, we review documented and emerging applications of LLMs across four main research tasks in fisheries science: literature review and knowledge synthesis, data collection and processing, developing and evaluating models, and science communication and stakeholder engagement. We then draw on selected examples from ecology, environmental science, and oceanography to identify potentially transferable applications and research gaps relevant to fisheries. Viewed in this cross-disciplinary context, fisheries have already begun to apply LLMs in data focused tasks, whereas applications for modeling support and stakeholder communication are still mostly at a conceptual stage. Second, we synthesize technical, operational, legal and compliance, and ethical and societal risks that are particularly relevant to fisheries, and outline practical options for mitigation. Based on this analysis, we propose a six-step, risk-aware workflow for incorporating LLMs into fisheries research and advisory processes as supporting tools rather than as autonomous decision makers. Our aim is to help fisheries scientists, managers and policymakers explore LLMs in a cautious and transparent way, so that the benefits to sustainable fisheries management are realized without undermining scientific integrity or trust.</p>