AI learns new information by a process called machine learning, which makes it find better answers with time through data analysis. There exists supervised learning, where the AI goes through labelled data where answers are known, and unsupervised learning, where AI finds patterns in data based on a label that was already determined.
Reinforcement learning is where AI learns from the outcome of its actions, whether rewards or penalties. AI models work with algorithms in processing large amounts of data in order to predict and adapt its behavior for high accuracy, and often learn in an iterative loop from new experiences.