Abstract

As generative AI systems like ChatGPT increasingly mediate product evaluation, understanding how they respond to variations in product descriptions is critical. This study investigates GPT-generated preference ratings across two product types, sunscreen (consumable) and speaker (non-consumable), using a 2 (Length: Long vs. Short) × 2 (Quality: High vs. Low) within-subjects design. Our results show that both longer and higher-quality descriptions receive higher ratings. We also find support that the impact of longer descriptions is stronger for lower quality products. As GPT begins to shape consumer behaviour, our results suggest that product content may increasingly be optimised for LLMs, potentially diverging from human preferences.

Share

COinS