Background: Most bladder cancer cases present as non-muscle-invasive bladder cancer (NMIBC), with the course of multiple recurrences leading to stage progression to muscle-invasive bladder cancer (MIBC) in 10–20% of cases, which is associated with higher morbidity and mortality. Accurate histopathologic classification of bladder cancer remains important for patient management. Methods: This retrospective–prospective observational cohort study was conducted on 244 transurethral resection specimens. Immunohistochemistry assessed CK5/6, CK20, and apoptosis-inducing factor (AIF) using three representations: intensity, percentage of positive cells, and multiplicative score. Discrimination between NMIBC (pTa/pT1) and MIBC (≥pT2), and between low-grade (LG) and high-grade (HG) tumors, was evaluated using ROC/AUC analysis and logistic regression. The main analysis focused on cross-sectional marker performance in primary/non-recurrent tumors. Recurrent tumors were analyzed only as an exploratory subgroup. Tumors were also categorized into basal, luminal, mixed/double-positive, and double-negative phenotypes using thresholds of 10% for CK5/6 and CK20. Results: For stage discrimination, all three markers showed modest separation. The best-performing representation was CK5/6 intensity (AUC 0.641; lower in MIBC). For grade discrimination, the AIF score showed the highest performance (AUC 0.729, higher in HG). Combining markers improved model performance (NMIBC vs. MIBC: AUC 0.784; strict LG vs. HG: AUC 0.778). Using the 10% cutoff in non-recurrent tumors, mixed/double-positive tumors had the lowest MIBC proportion (6.0%) and double-negative tumors the highest (46.7%). Conclusions: CK5/6, CK20, and AIF provide modest discrimination between stages, with lower CK5/6 and CK20, and higher AIF, in MIBC. The AIF score shows the highest separation between grades and may serve as a useful non-proliferation marker for grading, particularly when interpreted alongside CK5/6 and CK20 in a simple immunohistochemical panel.