Brain tumours are one of the severe health issues that require an accurate diagnosis and identification, especially with the aid of an MRI. Using the classical image processing techniques along with the classical ML, there are some limitations in accurately identifying the tumour regions in the complex MRI images that usually contain noise and uneven quality of images. AI has impacted multiple aspects of healthcare by providing sophisticated methods for diagnosis or even treatment plans. An AI method based on DL and ML architecture is presented in this study to locate brain tumours using MRI images. In light of multiple uses of DL, such as pre-processing, segmentation, feature extraction, classification, and evaluation, this article shall address the major DL concepts related to brain tumours. The suggested ML and DL methods for brain tumour diagnosis employing MR images are reviewed and evaluated in this study, which provides a modern evaluation. Cerebral neoplasms continue to present a major task for clinicians as they are often complex and come in many forms. It is crucial to begin screening as early as possible to ensure high accuracy in the differential diagnosis. This paper reviews various ML and DL frameworks such as CNNs, ensemble learning, transfer learning, and other hybrid models, among others. The model’s accuracy and performance are assessed and compared with benchmark datasets and measures. Lastly, the paper reveals the shortcomings of the methods utilised in the detection of brain tumours, including issues related to do with biases in data sets and computational costs, and possibilities for enhancement. At the same time, it examines the practical significance of the findings in question for clinical practice and points to the value of novel MRI-based diagnostic technologies for patients. Overall, this review provides valuable insights, highlights current trends, and outlines future directions for research in this important area of medical imaging analysis.