ChatGPT Scores Drop: Prompt Design Key to AI Quality

New data shows AI-generated business texts score low on quality and readability. On a 100-point scale, the average score was just 44.1, with 60 points considered acceptable. ChatGPT texts, for example, scored 37.7 points. What's interesting is that the study found prompt design, not the specific AI model, significantly impacts quality. Scores for the same model varied from 1 to 96 points based on how the prompt was written. A simple instruction like "write clearly" boosted readability to 79.4 points, though it reduced subject detail. This suggests that improving AI text quality might be an instruction-design problem, rather than a limitation of the AI itself.