This study described the development of an AI model, called UC-TIL, which had the goal of quantifying the spatial patterns of tumor-infiltrating lymphocytes(TIL) in cases of urothelial carcinoma in order to predict survival and immunotherapy response. This was a retrospective study that sourced information from 558 patient cases from public datasets, Emory Hospital, and Cleveland Clinic in order to train the model and evaluate its predictive ability. An AI tool called Hover-net classified the tumor tissue nuclei into 5 different categories(neoplastic, non-neoplastic epithelial, inflammatory, connective, and dead cells) and identified the "inflammatory" category as the one containing the TILs and then UC-TIL analyzed the spatial arrangements of these TILs within these images. UC-TIL was able to accurately predict patient survival based on TIL images and completely independent from other clinical factors that may have been involved. Additionally, for patients receiving immune checkpoint inhibitor(ICI) therapy, UC-TIL was able to accurately predict which ones would be the best responders to the treatment and would thus receive the best outcome from it. UC-TIL was also able to outperform three pathologists in analyzing images and predicting survival chances from them. The scoring by the pathologists did not show significant association with survival but the scoring by UC-TIL did. The ability of UC-TIL to predict survival rates allows clinicians to stratify patients into low and high risk categories, and then tailor treatment plans around risk, which could improve outcomes for all patients.