For the first time, the in silico design, screening, and in vitro validation of potent GSK-3β type-II inhibitors are presented. In the absence of crystallographic evidence for a DFG-out GSK-3β activation loop conformation, computational models were designed using an adapted DOLPHIN approach and a method consisting of Prime loop refinement, induced-fit docking, and molecular dynamics. Virtual screening of the Biogenics subset from the ZINC database led to an initial selection of 20 Phase I compounds revealing two low micromolar inhibitors in an isolated enzyme assay. Twenty more analogues (Phase II compounds) related to the hit [pyrimidin-2-yl]amino-furo[3,2-b]furyl-urea scaffold were selected for structure-activity relationship analysis. The Phase II studies led to five highly potent nanomolar inhibitors, with compound 23 (IC =0.087 μM) > 100 times more potent than the best Phase I inhibitor, and selectivity for GSK-3β inhibition compared to homologous kinases was observed. Ex vivo experiments (SH-SY5Y cell lines) for tau hyperphosphorylation revealed promising neuroprotective effects at low micromolar concentrations. The type-II inhibitor design has been unraveled as a potential route toward more clinically effective GSK-3β inhibitors.