Investigating fake financial news: a systematic review of trends, themes, LLM-based detection methods, and theories

Purpose This paper conducts a systematic literature review (SLR) to explore research directions in fake financial news, addressing a growing concern in today’s digital financial landscape. Design/methodology/approach The paper applies the search, appraisal, synthesis, and analysis (SALSA) framework, utilizing nine databases and reviewing 40 studies from 2010–2025 (March). Findings This study unveils major trends, themes, state-of-the-art detection methods, and theoretical foundations in fake financial news research. It identifies significant variables such as impact, timing, target, and responses, providing a deeper understanding of fake financial news. Moreover, this review highlights that despite advances in large language models (LLMs)- based detection, human-centered factors remain largely overlooked. By extracting empirical and analytical insights from prior research, this study proposes a framework that connects detection methods and behavioral theories to guide future research in human-centered, interdisciplinary approaches for financial misinformation detection. Research limitations/implications The study offers practical take-aways for misinformation detection and amelioration for various financial stakeholders, including investors, firms, and policymakers. It recommends integration of LLM tools for real-time misinformation detection, establishing financial fact-checking platforms, incorporating corporate responses, and increasing financial literacy education. The study also suggests that stronger regulatory interventions by Congress and the SEC are essential to mitigating the threat of financial misinformation. Originality/value This paper contributes to the field of information management by synthesizing the current knowledge on fake financial news. It is among the frontier systematic reviews, integrating research trends, themes, theoretical foundations, and LLM-based detection advancements beyond prior computational approaches.