Human-Centred Generative AI for Affordable Housing Design in Nigeria Through an Architect-led Lifecycle Framework

Affordable housing in Nigeria requires architectural solutions that combine spatial efficiency, climate responsiveness, buildability, adaptability, maintainability and affordability in use. This critical structured narrative review examines how generative artificial intelligence (GenAI) can support architect-led but multidisciplinary affordable-housing design when combined with Building Information Modelling (BIM), conventional building-performance evaluation and post-occupancy learning, including digital-twin functions where technically and economically justified. Searches completed on 30 July 2026 used Scopus and Google Scholar for principal discovery, supplemented by targeted publisher-platform and institutional searches. The substantive analytical corpus comprises 43 peer-reviewed publications; review-methodology papers and policy, standards, legal and governance sources are interpreted separately. Evidence was coded across GenAI-specific option generation, conventional generative/parametric design, spatial efficiency, unit/block/site design, BIM information coordination, building-performance simulation, post-occupancy evaluation (POE), digital-twin applications, affordability, liveability and responsible AI. The synthesis indicates that GenAI is most defensible as a constrained exploratory option-generation layer whose outputs require independent architectural, statutory, performance and constructability verification; BIM supports information coordination and traceability rather than automatic validation; and POE provides the baseline lifecycle-learning mechanism, with digital twins representing an optional higher-maturity layer. Human-centredness is operationalised through household and resident evidence at briefing, option evaluation and post-occupancy stages, while accountable professional judgement remains explicit. The article proposes a six-stage lifecycle framework, Nigerian implementation-readiness conditions, an illustrative comparative pilot scenario and testable indicators for empirical validation against conventional design workflows.