AI chatbots have become the dominant gateway to scientific information, with platforms like OpenAI and Anthropic scraping thousands of pages for every visitor they send back to publishers. While academia focuses on AI's role in writing, the greater threat lies in how these systems silently distort knowledge discovery. AI recommendations systematically amplify citation bias, overrepresenting highly-cited papers and certain demographics while marginalizing innovative work. As these tools increasingly shape literature reviews and expert identification, the academic community urgently needs guidelines for responsible AI-assisted research. Without intervention, we risk creating a narrower, more homogenous scientific discourse that reinforces existing hierarchies rather than advancing knowledge.
Zhicheng Lin (Sun,) studied this question.