Generalization bias in large language model summarization of scientific research

Checked 5 Nov 2026. Open license (CC-BY, CC-BY-SA, CC0, public domain): full text indexed and used in synthesis.

Artificial intelligence chatbots driven by large language models (LLMs) have the potential to increase public science literacy and support scientific research, as they can quickly summarize complex scientific information in accessible terms. However, when summarizing scientific texts, LLMs may omit details that limit the scope of research conclusions, leading to generalizations of results broader than warranted by the original study. We tested 10 prominent LLMs, including ChatGPT-4o, ChatGPT-4.5, DeepSeek, LLaMA 3.3 70B, and Claude 3.7 Sonnet, comparing 4900 LLM-generated summaries to their original scientific texts. Even when explicitly prompted for accuracy, most LLMs produced broader generalizations of scientific results than those in the original texts, with DeepSeek, ChatGPT-4o, and LLaMA 3.3 70B overgeneralizing in 26–73% of cases. In a direct comparison of LLM-generated and human-authored science summaries, LLM summaries were nearly five times more likely to contain broad generalizations (odds ratio = 4.85, 95% CI [3.06, 7.70], p < 0.001). Notably, newer models tended to perform worse in generalization accuracy than earlier ones. Our results indicate a strong bias in many widely used LLMs towards overgeneralizing scientific conclusions, posing a significant risk of large-scale misinterpretations of research findings. We highlight potential mitigation strategies, including lowering LLM temperature settings and benchmarking LLMs for generalization accuracy.

• Messeri & Crockett 2024's warning that AI could let science "produce more but understand less" now has systematic empirical backing: Peters & Chin-Yee 2025 tested 10 prominent LLMs on summarizing papers from top journals and found a strong bias toward overgeneralizing scientific conclusions by omitting qualifiers, with newer models performing worse than earlier ones. Trace →
• The illusion of understanding that Messeri & Crockett 2024 warn could define a new phase of science now has a measured mechanism: Peters & Chin-Yee 2025 find that widely used LLMs systematically overgeneralize scientific conclusions when summarizing—explicitly risking illusions of understanding—and that newer models do this worse than earlier ones. Trace →
• Two independent benchmarking studies undercut the assumption that newer model generations are becoming epistemically safer: Linardon et al. 2025 found no clear evidence that GPT-4o reduces citation hallucination relative to earlier versions, and Peters & Chin-Yee 2025 found newer models generalize less accurately than their predecessors. Trace →
• The model-generation contingency offers no comfort about upgrades: Linardon et al. 2025 find no clear evidence that GPT-4o reduced citation hallucination, and Peters & Chin-Yee 2025 find newer models overgeneralize scientific conclusions worse than earlier ones. Newer generations cannot be assumed more reliable, so verification duties should not relax as models improve. Trace →

• Research integrity and LLMs