Diagnosing soil bioremediation failure: evidence-weighted pathways from mechanism to field closure.

Soil bioremediation is often presented as a sustainable alternative to physicochemical remediation, yet its field performance remains less predictable than laboratory and microcosm evidence suggests. This critical review argues that the central problem is not a lack of advanced technologies, but the frequent mismatch between the diagnosed failure mode, the selected intervention, and the endpoint used to claim success. Organic pollutants may be degraded, transformed, mineralized, or converted to hazardous intermediates, whereas metals and metalloids can only be immobilized, stabilized, extracted, accumulated, or shifted in speciation and bioavailability. Soils contaminated with per- and polyfluoroalkyl substances (PFAS)further expose the need to distinguish destruction from sorption, retention, leaching control, and exposure interruption. The review evaluates soil bioremediation through explicit criteria: mechanistic plausibility, soil realism, endpoint relevance, biological activity, temporal durability, field readiness, safety, scalability, and residual risk. The evidence indicates that underperformance commonly arises from mass-transfer limitation, contaminant aging, redox mismatch, oxygen and moisture heterogeneity, co-contamination, toxic intermediate accumulation, inoculant establishment failure, phytoremediation endpoint mismatch, and rebound after active treatment stops. Emerging tools are treated here as conditional evidence or conditional interventions, not as generic solutions. Nanomaterials are admissible as bioremediation only when they support a measured biological process rather than only sorption, catalysis, or immobilization. Omics and isotope-linked methods can strengthen causal inference, but closure still requires convergence with contaminant kinetics, intermediate control, toxicity reduction, and post-treatment stability. Synthetic microbial communities and engineered microbial systems must retain function after ecological filtering. Artificial intelligence and digital twins can structure decisions only within validated data domains and declared uncertainty bounds. The principal contribution of this review is a field-facing decision sequence that links failure mode, limiting mechanism, diagnostic evidence, admissible intervention, secondary-risk control, and closure. This framework shifts soil bioremediation from technology optimism toward diagnosis-led, evidence-weighted, risk-controlled deployment.