Solid-state batteries could be safer and more energy-dense than today’s lithium-ion technology, but finding materials that allow ions to move quickly through solid electrolytes has been difficult. Researchers developed a machine learning pipeline that predicts Raman spectra and identifies a distinctive low-frequency signal linked to liquid-like ion motion inside crystals. This signal appears when rapid ion movement temporarily disrupts a crystal’s symmetry. The approach could dramatically speed up the discovery of superionic materials for advanced batteries.