The article explores a method called prior swapping, which allows for quickly transforming inference results from one prior to another. This method can efficiently generate accurate posterior samples under different target priors, even if the initial prior used was incorrect. By using prior swapping, researchers can apply less costly inference algorithms to certain models and incorporate new or updated prior information after the initial inference. The study shows that importance sampling may not always work for this task, depending on the similarity between the false and target priors. The proposed method has theoretical guarantees and has been successfully tested on various models and priors.