Digital forensic analyses that are conducted manually using forensic tools can be laborious and error prone. The situation is exacerbated given the large volumes of digital evidence that have to be processed and analyzed. Large language models developed by state-of-the-art artificial intelligence research can be engaged to reduce the workloads of digital forensic practitioners. This chapter describes a framework that leverages large language models to automate chat log forensics. A novel metric is specified for assessing the forensic effectiveness of large language models from three key perspectives: factual evidence, criminal motive and subject relationships. The experimental results demonstrate the utility and performance of the framework in chat log investigations.