Back to today’s briefing

Story perspectives

ChatGPT Misleads on Retractions in Academic Research

9/24/2025

23 4 Full Breakdown

Summary

Story summary

Generative AI, including ChatGPT, is increasingly used for academic tasks like literature reviews.
In a study by Er-Te Zheng and Mike Thelwall, ChatGPT failed to recognize retracted articles.
ChatGPT praised nearly 75% of the flawed articles and did not mention their retractions.
In a separate test, ChatGPT confirmed claims from retracted articles as true in about two-thirds of cases.

History

Historical Context

The American Psychological Association's 2012 guidelines on ethical research practices underscored the necessity of accurate citation in academia. A 2023 study by Zheng and Thelwall revealed that ChatGPT misidentified nearly 75% of flawed articles, raising concerns about AI's reliability in research. This reflects ongoing debates about the integrity of research methods, echoing historical challenges faced by emerging technologies in academia.

1 of 2

Story summary

Generative AI, including ChatGPT, is increasingly used for academic tasks like literature reviews.
In a study by Er-Te Zheng and Mike Thelwall, ChatGPT failed to recognize retracted articles.
ChatGPT praised nearly 75% of the flawed articles and did not mention their retractions.
In a separate test, ChatGPT confirmed claims from retracted articles as true in about two-thirds of cases.

1 / 2