European Journal of Sustainable Agroecosystems

Background: Traditional agricultural extension services, which rely predominantly on in-person agents and static informational materials, increasingly struggle to deliver timely, personalized guidance to a growing and diverse global farming population, particularly in remote and resource-constrained regions. The recent emergence of Large Language Models (LLMs) and generative artificial intelligence offers a potential pathway toward scalable, natural-language agricultural advisory services. 
Objective: This review synthesizes current scientific and technical literature on LLM-assisted intelligent advisory systems for farm management, examining how these systems integrate multi-source agricultural data with natural language reasoning to support crop, irrigation, nutrient, and pest management decisions. 
Framework: Drawing on peer-reviewed and archival literature published primarily between 2020 and 2026, this review conceptually describes the architecture of LLM-based advisory frameworks, encompassing agricultural knowledge integration, retrieval-augmented reasoning, context-aware recommendation generation, explainability mechanisms, and iterative refinement through user feedback. 
Findings: The synthesized literature indicates that deployed and prototype LLM-assisted advisory platforms have been applied to crop selection, variety recommendation, irrigation scheduling, nutrient management, pest and disease diagnosis, and climate-adaptive farm planning, with several field-deployed systems reporting engagement across tens of thousands of farmers and hundreds of thousands of queries. Comparative and qualitative assessments in the literature consistently describe LLM-assisted systems as improving accessibility, personalization, multilingual reach, and scalability relative to conventional agent-based extension, while explainability, factual reliability, and reasoning depth remain comparatively less mature and are identified as active research priorities. 
Significance: This review provides a structured, literature-grounded synthesis relevant to researchers, agricultural technology developers, extension agencies, and policymakers seeking to understand the current evidence base for LLM-assisted farm advisory systems.
Conclusion: LLM-assisted advisory systems represent a scientifically credible and rapidly evolving pathway toward scalable, personalized, and increasingly explainable agricultural decision support, though realizing their full potential will require continued progress in factual grounding, domain-specific reasoning, and equitable access for smallholder and low-connectivity farming communities.