Evolution and adaptation of artificial intelligence (AI) algorithms in landscape architecture planning and design

The rapid evolution of artificial intelligence (AI) technology is driving a transformation in landscape architecture planning and design, shifting from experience-driven approaches to algorithm-enabled solutions. As AI has progressed from early expert systems to contemporary generative AI, its potential and advantages in addressing complex landscape architecture challenges have become increasingly evident. This paper synthesizes domestic and international literature and industry practices, focusing on four core capabilities, i.e. identification, optimization, generation, and understanding, to trace the development, current applications, and suitability of various algorithms in site analysis and evaluation, schematic design, visualization, and knowledge management. Generative algorithms, particularly AI-generated content (AIGC), demonstrate high applicability in scheme design and rapid rendering. Discriminative and optimization algorithms, meanwhile, hold advantages in assessment and optimization phases. However, AI still struggles with core challenges, including accounting for site-specific uniqueness, quantifying qualitative factors, simulating dynamic systems, representing complex human experiences and aesthetic values, producing high-precision design solutions, and obtaining sufficient high-quality data. The integration of AI has not only reshaped traditional design processes but also intensified tensions among AI algorithms, design problems, and designers. The black-box nature of AI algorithms complicates the assignment of control and responsibility for design decisions, the pace of technological iteration outstrips designers’ knowledge updates, specialized datasets for landscape architecture remain scarce, rapid generation undermines designers’ deep thinking, and it drives designers to transition from executors to guides, integrators, and evaluators. Future efforts should concentrate on three areas: technology development, problem constraints, and designer agency. Technologically, we should strengthen AIGC, multimodal, and hybrid intelligent systems, develop explainable AI and AI agents, enhance decision-making transparency and controllability, and promote efficient human-AI co-creation under conditions of mutual trust. In terms of problem constraints, we should integrate planning and design algorithms and use techniques like small-sample optimization and knowledge enhancement to improve AI’s adaptability to landscape architecture issues. In terms of designer agency, it is important to emphasize designers’ core value, reasonably allocate tasks between AI and designers, build a shared database encompassing local knowledge and public perception, advance the education system of landscape architecture reforms, and expand capability recognition and industry standard development.