The fracture resistance of structures is critically compromised by random manufacturing defects, such as micro-pores or micro-cracks, which act as preferential sites for crack nucleation. Conventional deterministic topology optimization ignores such defect uncertainties, often yielding fragile designs. To address this gap, we develop a stochastic topology optimization framework that explicitly incorporates random defect distributions to enhance fracture robustness. The objective maximizes the mean of total fracture energy associated with crack evolution under a fixed volume fraction. These results are estimated via Monte Carlo simulations coupling the phase-field fracture model with the Solid Isotropic Material with Penalization (SIMP) method. Sensitivities are computed using an adjoint formulation and aggregated across defect realizations, enabling gradient-based optimization via GCMMA or OC algorithms. The framework is validated through several benchmark problems involving diverse geometries and defect types (e.g., point-like pores and line cracks). Numerical results consistently demonstrate that stochastic designs achieve up to 50% higher mean fracture resistance and a 30-40% reduction in performance variability compared to deterministic solutions. Optimized topologies feature targeted reinforcement in statistically vulnerable regions, a feature absent in deterministic topology optimization. This work demonstrates that embedding stochastic defect modeling directly into topology optimization significantly improves both robustness and reliability in fracture-critical engineering applications.