OBJECTIVE: This study aimed to explore the causal effects of lipidome on Alzheimer's disease (AD) and the mediated effects of the metabolites using Mendelian randomization (MR). METHODS: Data were obtained in genome-wide association studies, and single-nucleotide polymorphisms were screened according to the underlying assumptions of MR. Subsequently, weighted inverse variance was used to analyze the causality of lipidome with AD as well as the mediated effects of metabolites. Finally, MR-Egger, Cochran's Q, and sensitivity analysis were used to assess horizontal pleiotropy, heterogeneity, and robustness of the results, respectively. RESULTS: The MR analysis showed that phosphatidylcholine (PC) (15:0_18:2) (mediated proportion: 18.30%, p = 0.024) and phosphatidylethanolamine (PE) (18:0_18:2) (mediated proportion: 14.60%, p = 0.028) mediated the reduction of AD risk by lowering betaine levels, which revealed lipidomic susceptibility. The MR-Egger intercept showed no horizontal pleiotropy for all results (p ≥ 0.05). Cochran's Q showed heterogeneity in some of the results. Sensitivity analysis indicated that all results were robust. CONCLUSION: Our findings reveal the pathways through which PC (15:0_18:2) and PE (18:0_18:2) mediated the reduction of AD risk by lowering betaine levels.