Artificial intelligence for cleaner water and climate-resilient water resource management: A systematic review of water quality, risk, and decision support

Artificial intelligence (AI) applications in water management have expanded rapidly, yet evidence remains fragmented across technological capability, methodological validity, climate robustness, sustainability, and progression from prediction to demonstrated water-system outcomes. This systematic review synthesized recent evidence on AI approaches, water-management domains, evidentiary progression, methodological constraints, and research priorities. Scopus, Web of Science Core Collection, and ScienceDirect were searched for English-language journal articles published January 2021–30 June 2026 using prespecified procedures. Of 312 records, 258 remained after duplicate removal, 51 underwent full-text assessment, and 24 met all eligibility criteria. Evidence was synthesized using qualitative thematic analysis, study-level methodological appraisal, evidence-stage mapping, and audits of climate robustness, validation independence, and quantitative sustainability. Analytical/predictive evidence occurred in all 24 publications (100.0%), environmental-intelligence evidence in 21/24 (87.5%), decision utility in 10/24 (41.7%), and operational-implementation relevance in 2/24 (8.3%); none demonstrated AI-attributable environmental or resource-management outcomes. Only 4/24 (16.7%) contained original empirical AI/model/tool evaluations. Climate robustness was limited: temporal-shift testing occurred in 2/24 (8.3%), spatial transfer/external validation in 1/24 (4.2%), while formal climate/hydrological perturbation testing was absent. Explicit quantitative sustainability assessment was absent in all 24 publications. Overall, the evidence supports analytical and decision-support capability but does not establish independent, transferable, climate-robust, quantitatively sustainable environmental outcomes. Key gaps are independent validation, transferability, causal attribution, operational evaluation, climate stress testing, and quantitative sustainability assessment. The AI-enabled Closed-loop Water Intelligence and Climate-resilience (AI-CWIC) framework is proposed as an author-derived, hypothesis-generating conceptual synthesis of these gaps, not an empirically validated causal or scoring model.