Artificial Intelligence Readiness of Bacterial Self-Healing Cement-Based Materials: Evidence, Design Constraints, and Research Priorities

Bacterial self-healing cement-based materials (BSHCMs) couple microbial mineralization with cement-based material design to autonomously seal cracks and potentially restore durability. However, their performance depends on a complex interaction among bacterial viability and physiological state, mineralization pathway, carrier and nutrient systems, calcium availability, matrix chemistry, crack characteristics, moisture, and exposure history. This review, supported by bibliometric mapping of 805 Scopus-indexed records, examines these interdependencies through the specific lens of artificial intelligence (AI) readiness. The mapping revealed six interconnected research themes spanning bacterial mineralization, sustainable cement-based systems, matrix chemistry and transport, encapsulation, durability, and machine learning (ML). AI evidence was classified as directly demonstrated in bacterial self-healing systems, transferable from adjacent concrete and structural health monitoring applications, or prospective. Although existing ML studies report high internal predictive performance, their engineering generalizability remains limited by heterogeneous and frequently literature-derived datasets, random train–test partitioning, potential feature leakage, synthetic data dependence, insufficient uncertainty reporting, and scarce independent laboratory or field validation. The analysis further shows that sustainability and economic benefits cannot be assumed from the biological nature of the technology but must be demonstrated through service life extension relative to the additional burdens of cultivation, nutrients, carriers, and processing. Advancing BSHCMs toward trustworthy AI-supported engineering, therefore, requires harmonized and machine-readable datasets, matched attribution controls, delayed cracking and realistic exposure experiments, explicit uncertainty and negative result reporting, study-grouped and external validation, and pilot- to field-scale testing. AI should consequently be regarded not as a substitute for biological healing, but as a decision support layer whose value depends fundamentally on the quality, traceability, and transferability of the underlying experimental evidence.