The rising prevalence of Type 2 diabetes presents a significant global health challenge that requires early detection and effective management. While machine learning models offer strong predictive capabilities for diabetes risk assessment, their lack of interpretability limits trust and clinical adoption. This paper proposes an explainability-driven prediction framework that integrates two complementary techniques: (1) SHAP-based feature attribution, enhanced through percentage-based normalization for better accessibility, and (2) personalized counterfactual explanations that generate actionable “what-if” scenarios tailored to individual user profiles. Unlike prior work that focuses solely on feature importance or abstract visualizations, our framework delivers both justification (“why” a prediction was made) and guidance (“how” risk can be reduced) through natural language insights and medically aligned recommendations. The framework is evaluated on a public dataset, achieving 96.7% accuracy. A user study involving general users and healthcare professionals demonstrates high satisfaction with the clarity, usefulness, and actionability of the explanations. Results show that integrating SHAP and counterfactual reasoning provides both justification and guidance, advancing explainable AI adoption in healthcare.