Artificial intelligence is increasingly interwoven with design thinking (DT), yet comparative, stage-by-stage syntheses across canonical DT models remain scarce. This literature review maps how AI augments and challenges the major stages of widely used models and relates these effects to five illustrative domains. Following the SPAR-4-SLR protocol, we searched the Web of Science (2005–August 2025), screened records in two stages and assembled a corpus of 205 eligible studies for comparative synthesis. Across models, AI scales early-stage evidence work through large-N text and behavioral analytics, widens ideation via generative systems and accelerates prototyping and testing through simulation and predictive evaluation; at the same time, risks include bias, privacy and sovereignty concerns, evaluation opacity and homogenization of creative output. The weight of evidence supports hybrid intelligence: allocate divergent exploration primarily to AI while retaining human judgment for convergent selection and ethical decision-making. A complementary AI-native “Stingray” model highlights concurrent train–develop–iterate workflows that treat AI as a co-designer, while underscoring governance needs around interpretability and auditability. Overall, the review offers a model-by-model, stage-specific map of AI’s roles in DT, along with practical guidance for responsible deployment and research priorities for assessing boundary conditions and external validity.