The expanding application of artificial intelligence in architecture offers opportunities to enhance design quality, improve decision-making, increase productivity, and support sustainable urban development. However, its effective and responsible use requires a coherent policy framework addressing legal, professional, institutional, technological, and human dimensions within the national context. This study aimed to identify key components and develop an AI-based policy framework for the architectural profession in Iran. The research is developmental in nature and employs a qualitative, exploratory approach conducted in two stages. First, a qualitative meta-synthesis was conducted in accordance with PRISMA 2020 guidelines to identify and organize preliminary policy components. Of 2,018 records identified,60 studies were selected for final analysis following screening and quality assessment. Second, the extracted components were evaluated by 20 experts through three Delphi rounds based on necessity, importance, contextual suitability, and feasibility, and validated using the Content Validity Ratio and Content Validity Index . Six main components were identified: regulation and professional responsibility; data governance and infrastructure; ethics and transparency; human capital and education; institutional coordination and stakeholder engagement; and implementation and evaluation.The findings indicate that effective AI policymaking in architecture requires alignment among legal and professional requirements, data infrastructure, ethical principles, human capacities, institutional coordination, and monitoring mechanisms. Accordingly, the final framework was structured at strategic, institutional, and professional levels. It provides a basis for developing specialized regulations and standards, strengthening human and data capacities, defining professional responsibilities, and establishing implementation and evaluation mechanisms for AI in the Iranian architectural profession supporting sustainable urban development.