The editors of the leading journal Organization Science conducted an internal investiga... — 21 мая 2026 г. в 11:45:33.120
The editors of the leading journal Organization Science conducted an internal investigation and came to sobering conclusions. Since the launch of ChatGPT, the number of manuscript submissions has increased by 42%, but the quality of scientific prose has declined. This wave was driven by AI. 🧠 Study Summary A dedicated journal working group analyzed nearly 7,000 manuscripts and 10,000 reviews over the past five years using the Pangram detector. Their goal was to understand how generative AI has impacted the submission and review process since the launch of ChatGPT in November 2022. 📊 Key Findings 1. A Sharp Increase in Submissions Thanks to AI Since the launch of ChatGPT, the number of manuscript submissions has increased by 42% compared to the previous period. Almost all of this growth was driven by AI-powered texts. 2. Decline in Text Quality with AI The subjective feeling of "emptiness" in text was confirmed by the data: the readability of AI-authored texts is significantly worse (a 1.28 standard deviation drop in the Flesch Reading Index). Such texts are more difficult to comprehend. Authors also lose vocabulary and style diversity, creating a sterile but impoverished language. 3. The Level of AI Use Determines the Fate of an Article The researchers divided articles into groups based on the percentage of AI-authored text. Result: ▪️Low AI (0-30%): desk rejection probability of 44%. ▪️High AI (>70%): desk rejection rate reaches 70%. ▪️Publication Probability: 12% for articles with low AI versus 3.2% for articles with high AI. 4. Most alarming: AI has infiltrated peer review. Reviewers have also begun using AI to write reviews, especially in the 30-70% AI-powered text segment. Characteristics of such reviews: Less readable (as with papers) and monotonous – focusing on theory rather than empirical evidence, missing nuances. The authors emphasize that the problem isn't AI itself, but its combination with the "publish or perish" mentality. Business schools, where quantity is valued, have strong incentives to produce articles. AI has simply become a tool for satisfying these incentives – a rational response to distorted incentives. 💬 What do you think? Have you encountered "AI-style" in scientific texts or reports? Do you consider this a problem or a natural evolution?