AI Review Grades Science Claims Across 11 Fields
A survey by researchers at 94 institutions assigns validation grades to AI-driven research claims in each field.
Updated on Nov. 18, 2026 in Artificial Intelligence

Live Poll
Should the U.S. prioritize AI research funding despite its rising electricity demand?
Researchers at 94 institutions published a review of AI-driven scientific research across 11 fields. The review assigns A-to-D validation grades to AI research claims in each field.
Why it matters
The review examines how much of the output from AI systems that read papers, propose hypotheses and run experiments can be trusted. The authors frame the next decade as the test of AI-driven science, with AI serving as a cognitive copilot rather than a replacement for scientists.
The review covers published work across 11 fields and grades AI research claims on an A-to-D scale, while global data centers used about 415 terawatt-hours of electricity in 2024. Whether AI-driven research claims hold up at the grades assigned remains an open question the authors raise.
The players
Yongjun Xu
Lead author affiliated with the Institute of Computing Technology at the Chinese Academy of Sciences, who co-led the review.
Heyang Li
Lead author affiliated with the University of Chinese Academy of Sciences, who co-led the review.
The details
The review describes AI systems built from six modules that feed each other, and some systems read papers, propose hypotheses, direct robots in the lab and adjust experiments based on results. An autonomous laboratory called A-Lab was handed 58 target compounds and made 41 of them in 17 days, drawing synthesis recipes from the literature and adjusting when a route failed.
Timeline
2021: The lead authors' earlier review described AI as a tool for analyzing data, with AlphaFold as the model case.
2024: Global data centers used about 415 terawatt-hours of electricity.
2030: The review projects global data center electricity use will roughly double.
Big Picture
The authors' 2021 review framed AlphaFold as the model case for AI as a data-analysis tool, and this review extends that framing toward AI systems that run the full research loop. The authors describe the next decade as the test of whether that loop produces reliable science.
For readers, the review's grading of AI research claims offers a way to judge how much weight to give headline findings from AI-driven studies. The projected doubling of data center electricity use by 2030 may also affect the energy costs and infrastructure of services that depend on AI.
The takeaway
Treat AI-generated research claims as provisional until their validation grade is known. Checking whether a claim is graded and how strongly supported it is can help separate reliable findings from speculative ones.
Further reading
Explore more on Artificial Intelligence to see how AI is being applied across scientific research.
Source note: This article includes information reported by Earth.
Live Poll
Should the U.S. prioritize AI research funding despite its rising electricity demand?



