Alzheimer’s disease (AD) is a complex neurodegenerative disorder characterised by cognitive decline, which poses significant challenges for diagnosis and treatment. This study aims to evaluate and compare the performance of various machine learning (ML) classifiers, including Automated ML (AutoML), in predicting AD biomarkers using gene expression data. The study utilised three brain tissue datasets from Gene Expression Omnibus (GEO) database, comprising a total of 445 samples, to identify key differentially expressed genes (DEGs) associated with AD. Data preprocessing included outlier detection, normalisation, and undersampling to address class imbalance. Feature selection was conducted using Boruta’s algorithm, and the top 30 DEGs were analysed. Seven ML classifiers—SVM, Naïve Bayes, KNN, LDA, Ridge, Lasso, and elastic net logistic regression along with AutoML—were applied to the dataset, with performance evaluated based on sensitivity, accuracy, specificity, and error rate. The AutoML approach outperformed other models, achieving 83.17% accuracy and a low error rate of 16.83% on the test set, while the least-performing model, Naïve Bayes, achieved only 56.25% accuracy with a high error rate of 43.75%. These findings underscore AutoML’s potential in optimising biomarkers prediction, advancing personalised medicine, and improving AD diagnosis.