Causal Discovery (CD) is the task of automatically inferring causal structures, typically from obser- vational data. Recently, there has been much interest in utilizing domain knowledge from large language models (LLM) in causal discovery. However, existing LLM-based approaches only output a single directed acyclic graph (DAG) without uncertainty, which can be unreliable. Thus, the focus of the present disserta- tion is exploring Bayesian Structure Learning algorithms which produce distributions of graphs, natural representations of uncertainty and using LLMs to enhance their inference power. The present disser- tation has three main goals. The first is discussing and reviewing the basic algorithms of CD, and also focusing on LLM-based approaches. There is also a short section that explores different methods of defin- ing the LLMs output reliability. Secondly, we intend to investigate the usage of LLMs alongside Bayesian structure learning (BSL) methods for causal discovery. In particular, we propose to harness the domain knowledge from the LLMs as the prior distribution over graphs. Our experiments show that LLM-informed priors can improve the performance of BSL methods. Lastly, this text describes another approach, which focuses on Ancestral Graphs (AGs), where we do not need the sufficiency assumption. Furthermore, our model describes a method that uses the LLM as an expert-in-the-loop, mixing the data information with the outside knowledge during the model processing. Our experiments show that our method is competitive against others that address latent confounding on both synthetic and real-world datasets; and our design for incorporating feedback from a (simulated) human expert or a Large Language Model (LLM) improves inference quality.