Obesity is a major global public health issue linked to a wide range of chronic diseases. Understanding its complex causal pathways requires robust analytical methods. Mendelian randomization (MR), which employs genetic variants as instrumental variables, effectively addresses confounding and reverse causation and has become a key tool in obesity research. This review summarizes the development of MR methodologies, from single-sample to multivariable, mediation, and time-series models, and highlights key findings from the past decade. MR studies have revealed causal associations between obesity and nine major disease categories, including cardiovascular, metabolic, cancer, psychiatric, respiratory, renal, reproductive, musculoskeletal, and dermatological disorders. Obesity influences disease risk through mechanisms involving energy metabolism, hormonal regulation, and inflammation, with heterogeneity by age, sex, and fat distribution. Key genes such as , and have been identified as potential therapeutic targets. Current challenges include instrument strength, pleiotropy, population stratification, and the external validity of GWAS data. Future research that integrates multi-ancestry GWAS, functional validation, and multi-omics approaches may further enhance the utility of Mendelian randomization. MR provides a robust genetic framework for elucidating obesity's causal effects and informing targeted interventions and personalized treatment strategies.