Biotechnology Journal International

Precision agriculture is increasingly described as the convergence of two originally separate technological programmes: engineered nanomaterials that act at the plant and soil interface, and computational learning systems that convert agronomic data into management decisions. The claim that these programmes already constitute a single integrated capability has become common in recent literature, yet the evidential basis for that integration has rarely been examined with the scepticism it warrants. This critical narrative review evaluates the strength, consistency and methodological quality of the evidence linking artificial intelligence to plant nanotechnology, and asks where the coupling is demonstrated, where it is merely plausible and where it is rhetorical. Literature was identified through Crossref Metadata Search, PubMed, the Directory of Open Access Journals and targeted searching of publisher and institutional pages, supplemented by backward and forward citation tracing, with all bibliographic records verified through digital object identifier resolution. Four coupling modes are distinguished: nanoscale sensing that generates machine-readable plant signals, data-driven prediction and design of nanomaterial behaviour, stimuli-responsive delivery that actuates algorithmic decisions, and decision integration at field scale. Evidence quality differs sharply among these modes. Supervised models of nanoparticle uptake and plant response now rest on curated datasets and interpretable learning methods, but they inherit descriptor limitations from nano quantitative structure and activity relationship modelling, rely on small and heterogeneous laboratory datasets, and have seldom been validated prospectively. Nanosensors detect defined stress signalling molecules in living tissue with high temporal resolution, yet reports of calibration stability, cross-species transferability and field durability remain scarce. Nano-enabled fertilisers and pesticides show efficiency gains in controlled conditions that are frequently attenuated or unverified in field systems, and comparisons with conventional analogues are often methodologically weak. Environmental fate, soil microbiome effects and life-cycle burdens remain insufficiently characterised to support confident safe-by-design claims. The dominant limitation is not conceptual but infrastructural: the coupling of artificial intelligence to plant nanotechnology is constrained by data scarcity, non-standardised reporting and an unresolved gap between glasshouse demonstration and agronomic deployment. Research priorities are proposed that address prospective validation, standardised nano-agronomic datasets, field-durable sensing and governance of data asymmetry.