The spread of financial misinformation online poses a serious threat to markets and investors. Could AI be the solution? Researchers explored this question in the recent COLING 2025 Financial Misinformation Detection (FMD) Challenge. The challenge focused on building AI systems that could not only identify false financial claims but also explain their reasoning. This transparency is crucial for building trust in AI-driven financial analysis. The team experimented with several large language models (LLMs), including Qwen, Mistral, and Gemma-2, fine-tuning them on a dataset of financial claims labeled as true, false, or not enough information. They discovered that simply training an LLM to classify claims wasn't enough. The best results came from a sequential learning approach. They first trained the model to just classify the claims. Then, in a second stage, they trained it to generate explanations for its classifications. This two-step process, dubbed “SeQwen,” resulted in a significant boost in performance, especially in the quality and coherence of the explanations generated. SeQwen achieved an F1-score of 0.8283 for classification and a ROUGE-1 score of 0.7253 for explanations, outperforming models trained in a single phase. This research shows the potential of LLMs to combat financial misinformation and enhance transparency in financial analysis. However, challenges remain. Fine-tuning these powerful models requires significant computational resources. Also, evaluating the quality of explanations is complex and often requires human judgment. Despite these hurdles, the research demonstrates the exciting possibilities of AI for creating a more trustworthy and informed financial landscape. Future research will likely focus on improving the efficiency of training these models and developing even more sophisticated methods for evaluating their explanations.