The detection of hate speech is a difficult task and an active study subject at the moment. Effective hate speech identification is more important than ever in a time when a sizable section of the populace uses social media to express their emotions, feelings, rage, and anxiety as well as to look for protection. Conventional approaches to detecting hate content typically rely on hate-related keywords or utilize a labelled dataset to train a learning model. However, hate speech often stands out more due to its context. Conversely, training a learning model requires a substantial amount of labelled data. This paper explores a pioneering approach to hate content detection in the Bengali language through the innovative use of prompt engineering. We develop and apply dynamic prompts that improve the model’s comprehension and recognition of complex hate speech in a range of circumstances by utilizing the power of sophisticated language models. Using the publicly accessible HASOC 2023 dataset, we tested ChatGPT in our experiment and compared its performance to the state-of-the-art.