An increase in the volume of false information circulating as a direct consequence of the rise in the growth of social media has an effect of misguiding the general population. Therefore, a mechanism for identifying fake news is required to prevent such repercussions. Almost all of these already existing algorithms for detecting fake news work with resource-rich languages such as Spanish and English; however, few techniques can work for resource-constrained languages such as Urdu. The study aims to identify instances of fake news written in Urdu by applying deep learning and machine learning methodologies. We use the MuRIL and T5 models for the implementation process because these models were developed specifically for Urdu and Hindi language recognition. The newspaper articles included in the valid subset derived from credible news sources, and the accuracy of these items have been checked by hand. Inside the misinformation subsection, the problem of how challenging it was to discover fake news was overcome by employing experienced reporters who were native Urdu speakers and instructing them to compose deceptive news items purposefully. It has allowed the researchers to overcome the problem. The dataset covers various subjects, including business, sports, health, showbiz, and technologies. We have carried out baseline classification to use our Urdu database as a standard for other datasets. The experiments with the various systems have indicated that the MuRIL model significantly improves over the other models, such as the T5 Model, and attains an average F1 score of 0.96 and a validation accuracy of 0.83. These results have been determined based on the findings of the experiments.