Can AI pick the best watermelon? Examining Generative AI's effectiveness in fruit quality assessment

Traditionally, fruit quality assessment relies on sensory tests and laboratory analyses, such as spectroscopy, refractometry, and firmness tests. However, these methods are often time-consuming, expensive, and impractical for everyday consumers. The study proposes an AI-powered image recognition approach, using ChatGPT’s GPT-4o model to analyze watermelon images based on key external features.

Artificial intelligence (AI) is transforming industries, and horticulture is no exception. A new study, "Effectiveness of Generative AI Tool to Determine Fruit Quality: Watermelon Case Study," published in Horticulturae (2025), explores how generative AI models, specifically ChatGPT, can assess the quality of watermelons based on external characteristics. This research introduces a non-destructive, AI-driven approach that could enhance consumer decision-making and reduce fruit wastage.

Watermelon selection has always been a challenging and subjective process. Consumers rely on visual indicators such as color, striping, shape, and sound to judge ripeness and taste. However, these methods are often inconsistent, leading to dissatisfaction. The study investigates whether AI can outperform human intuition in selecting the best watermelon by analyzing images captured in retail environments.

How generative AI is used to assess fruit quality

Traditionally, fruit quality assessment relies on sensory tests and laboratory analyses, such as spectroscopy, refractometry, and firmness tests. However, these methods are often time-consuming, expensive, and impractical for everyday consumers. The study proposes an AI-powered image recognition approach, using ChatGPT’s GPT-4o model to analyze watermelon images based on key external features.

The study outlines five key steps in how AI assesses watermelon quality:

• Image Uploading: Consumers take smartphone photos of watermelons in a store.
• Feature Extraction: AI identifies color, shape, texture, and striping patterns.
• Analysis & Evaluation: AI compares the extracted features with known indicators of ripeness.
• AI Interpretation: The system provides a quality score and selects the best watermelon based on the dataset.
• User Feedback: Consumers validate the AI’s selection through sensory evaluation tests.

The AI model, based on convolutional neural networks (CNNs), mimics human perception but removes biases and inconsistencies, offering a more standardized and reliable selection process.

Comparing AI and human selection: Key findings

The study conducted two case studies to test whether AI’s watermelon selection aligns with human taste preferences. Evaluators - including farmers, food engineering students, and retailers - compared the AI-selected watermelons with sensory evaluations of sweetness, juiciness, crispiness, and freshness.

Below are the key findings:

• AI-selected watermelons were rated significantly higher in sensory attributes.
• Human testers preferred the AI-chosen watermelons in terms of taste and texture.
• AI was more consistent and unbiased, whereas human selection varied based on personal preferences and experience levels.
• AI assessments correlated with retailers’ preferences, supporting its commercial potential for improving fruit selection accuracy.

The study applied Wilcoxon rank sum tests and paired t-tests to validate the results. The p-values were below 0.05, indicating a statistically significant difference in quality perception between AI-selected and manually selected watermelons.

Challenges and future potential of AI in fruit quality assessment

While AI has demonstrated its potential to revolutionize fruit selection process, several challenges remain:

AI’s dependence on image quality

The AI model relies on high-quality images for accurate predictions. Variations in lighting, angles, and smartphone camera quality could impact results.

AI’s limitation in detecting internal quality

While AI effectively analyzes external traits, it cannot directly measure sweetness or crispiness - attributes traditionally assessed via sensory evaluation or lab tests. Future models may integrate spectroscopy or AI-enhanced ultrasound scanning for deeper insights.

Consumer trust and adoption

Many consumers still rely on traditional selection methods and may hesitate to trust AI recommendations. Increased awareness and real-world testing could help boost AI adoption in supermarkets and online grocery platforms.

Expanding AI’s use beyond watermelons

The success of AI in watermelon selection suggests potential applications for other fruits, such as mangoes, avocados, and peaches, where ripeness is difficult to assess externally. Future studies may refine AI models for multi-fruit assessments.

Future of smart agriculture and retail with AI

The findings from this study suggest that AI-powered fruit quality assessment could become an essential tool in both agriculture and retail. AI-driven selection could:

• Reduce food waste by ensuring only high-quality fruits reach consumers.
• Improve supply chain efficiency by optimizing fruit sorting and distribution.
• Enhance consumer confidence by providing objective, data-driven quality assessments.
• Support retailers and farmers in maximizing sales and reducing customer complaints.

As AI becomes more sophisticated and accessible, its integration into agriculture and food retail will likely expand, offering automated, accurate, and scalable solutions for fruit quality detection.