The conversion of CO₂ into gasoline-range hydrocarbons (C₅–C₁₂) using Fischer–Tropsch synthesis (FTS) represents a compelling pathway toward sustainable fuel production. In this study, we compiled and statistically analyzed a dataset of over 100 experimental records from the published literature focused on CO₂-FTS performance, predominantly featuring Co‑ and Fe‑based catalysts, which are the most frequently reported in gasoline-range studies. We evaluated four machine learning models—XGBoost, CatBoost, Random Forest, and Neural Networks—to predict CO₂ conversion and gasoline-range selectivity. CatBoost achieved the highest predictive accuracy with a test R² score of approximately 0.8, and was selected for further interpretation using SHAP-based post hoc analysis. The model revealed that the optimal operational conditions for maximizing gasoline-range hydrocarbon yield are aligned with ranges commonly reported: a temperature of 280–320 °C, pressure of around 2 MPa, and space velocity (GHSV) between 900 and 120,000 mL  h⁻¹ g⁻¹ (most studies cluster in the 1,000–5,000 range). Conditions were associated with enhanced chain growth probability and suppressed methane formation, especially in Co-based systems. The SHAP analysis also highlighted the principal role of catalysts containing cobalt (often supported on γ-Al₂O₃ with Re promoter) in increasing C₅⁺ chain growth and gasoline-range selectivity. Additionally, Co-based catalysts demonstrated clear benefits: increased chain-growth probability, reduced methane selectivity, and higher selectivity toward gasoline fractions under the identified optimal conditions. Our ML-driven framework not only predicts performance but also provides mechanistic insights into the influence of catalyst composition and reaction parameters. This integrated approach accelerates rational catalyst and process design for CO₂‑to‑fuel technology.