FROM PREDICTION TO PRACTICE: A BIBLIOMETRIC AND SYSTEMATIC REVIEW OF ARTIFICIAL INTELLIGENCE IN PRECISION HORTICULTURE—FIELD VALIDATION, CROSS CROP GENERALIZABILITY AND DECISION-SUPPORT READINESS

Abstract

Artificial intelligence is increasing the ability to monitor, analyse, and operate horticultural production offering opportunities in crop monitoring, resource allocation, and production efficiency. However, predictive performance in controlled experimental settings may not lead to reliable agronomic results or valuable into practice. This review examines three interconnected questions in the walk of transition from algorithmic development to operations: if intelligent systems are resilient in the face of varied production conditions, if knowledge and learning transfer effectively across crops and environments, and if model outputs provide defensible information management. Research examples of disease diagnostic, orchard monitoring, greenhouse control, and robotic harvesting enable us to put advances in context by considering current capabilities, methodological limits and practical adoption. There is also a focus on dataset representativeness, independent validation, environmental variability, uncertainty estimation and agronomic knowledge integration into decision processes. We outline a reproducible framework for the following bibliometric and systematic analysis, openly describing methods of literature search, study selection, evidence extraction and critical appraisal without claiming an actual database census. What is more important than individual metrics of predictive accuracy is that the predictor can meet a number of practical readiness requirements: consistent performance in all relevant settings, explicit communication of uncertainty of estimates, biological constraints, operational constraints and demonstrated benefits for growers. Future research should focus on transferable learning, field testing designs to avoid bias, clarity in its advice and comparing with current practices. Ultimately, the real test of progress in horticultural artificial intelligence is reliable decisions and tangible gains in productivity, resource efficiency and operational feasibility.