Background: Acetylcholinesterase (AChE) remains a clinically validated target for symptomatic treatment of Alzheimer’s disease (AD), yet current inhibitors display only moderate brain penetration and dose-limiting adverse effects. Ferulic acid (FA) exhibits neuroprotective properties and weak AChE inhibition, encouraging structural optimisation. Methods: The chemical structures and Ellman-assay IC₅₀ values for twenty FA derivatives were taken as reported in a previous external study. Three-dimensional geometries were energy-minimised (MM⁺→PM3) and analysed in Dragon 7 to generate 4,885 molecular descriptors. After variance/multicollinearity filtering, 918 descriptors remained. Stepwise feature selection identified four informative variables—nCL, SpMax2_Bh(p), Mor16s, SpMax7_Bh(s)—capturing halogen content, electronic distribution, and 3‑D shape. An artificial neural network (MLP 4‑8‑1, BFGS optimisation) was trained on a 14‑compound set and validated by leave‑one‑out cross‑validation and an external three‑compound test set. Results: The model reproduced experimental data with R² = 0.959, Q² = 0.956 and MAE = 1.43 µM; external prediction yielded R² = 0.927. Sensitivity analysis revealed SpMax2_Bh(p) as the principal driver of potency, while nCL highlighted the favourable impact of chloro‑substitution. Conclusions: A concise ANN‑QSAR model delivers accurate, mechanism‑based predictions of AChE inhibition for FA derivatives and offers clear design rules—enhanced π‑cloud polarisability and selective halogenation—for the next generation of brain‑penetrant AChE inhibitors. The workflow shortens the discovery timeline for AD therapeutics and can be expanded to larger libraries and multi‑target optimisation.