In recent years glucagon-like peptide receptor 1 (GLP-1R) has emerged as a 'hot target' for the treatment of insulin-related conditions. Small molecule positive allosteric modulators (PAMs) offer a promising alternative to peptide-based drugs currently on the market which suffer from poor oral bioavailability. PAMs act as molecular glues between GLP-1R and its natural substrate GLP-1, stabilising receptor–ligand interactions and boosting the receptor's response to natural GLP-1. After studying the dynamics of the GLP-1R/GLP1/PAM system, we identified suitable molecular dynamics snapshots and designed an active learning workflow to generate novel compounds and iteratively improve their properties. Using this workflow, we identified promising novel, diverse GLP-1R PAMs with potential for further development.