Evaluating AI Tools for Crop Management - What Farmers Should Know
Researchers at Penn State Extension are examining how artificial intelligence can assist farmers with decision-making, while also emphasizing the need for dependable, science-based tools.
A recent evaluation led by James R. Ladlee and horticulture educator Leah Fronk assessed eight large language model (LLM) platforms using ten practical questions focused on cucumber production.
The questions addressed essential topics such as pest and disease control, pesticide recommendations, nutrient management, and general crop practices.
Researchers compared AI-generated responses with established guidance from Extension publications and production manuals to measure both accuracy and usefulness.
The results revealed inconsistent performance across platforms. While some tools, including Gemini Free and Grok3, produced relatively strong and relevant answers, others offered incomplete or overly general responses.
In some cases, the information could be misleading, particularly when applied to specific regional conditions. These findings were shared at the Mid-Atlantic Fruit and Vegetable Convention in Hershey, Pennsylvania.
Although users are becoming more skilled at interacting with AI through improved prompts and follow-up questions, the study highlighted that general-purpose tools still lack reliability for agriculture-specific decisions. This has reinforced the importance of developing systems grounded in localized and research-backed data.
To meet this need, Penn State Extension introduced Tilva in early 2026. This specialized AI platform delivers responses based only on vetted Extension materials and government data, including resources from the Pennsylvania Department of Agriculture and the USDA. It also supports both English and Spanish users, improving accessibility.
Tilva includes a feedback feature that allows users to rate responses and provide input, helping refine its performance over time. Despite these advancements, experts stress that major farm decisions should still involve professional consultation.
Learn more about Tilva: https://extension.psu.edu/ai-assistant.
Photo Credit: penn-state-extension