Subsymbolic AI refers to AI approaches that represent knowledge through distributed patterns, learned weights, or numeric structures rather than explicit human-readable rules. Neural networks and deep learning are common examples, emphasizing data-driven pattern recognition over symbolic reasoning.