Researchers developed a machine-learning algorithm to accurately predict free sugar content in packaged foods globally. The model enables large-scale assessment of carbohydrate quality based on fiber and sugar content.