Scientific publishing is growing faster than any individual researcher can realistically keep up with, and the backlog is creating a real problem. Systematic reviews—among the most rigorous forms of evidence synthesis—can take months or even years to complete, raising the risk that important findings go unnoticed simply because there is too much literature to process. A new AI-assisted platform called MetaBeeAI was developed to address that gap.
MetaBeeAI combines large language models with human validation to help researchers automatically extract, organize, benchmark, and analyze information from hundreds or thousands of full-text scientific papers, while maintaining traceability and scientific rigor. The platform identifies relevant information within papers using AI, then allows experts to verify, correct, and refine outputs through a transparent review interface linked directly to the original source text.
"The system is not designed to replace scientists. Instead, MetaBeeAI aims to support researchers by reducing repetitive workloads and helping experts focus on interpretation, decision-making, and scientific insight," said Rachel Parkinson, lead author on the research published in Ecological Informatics.
The system was tested using nearly 1,000 scientific papers on pesticides and bees. It performed particularly well on factual extraction tasks, such as identifying bee species and pesticide compounds. Iterative expert feedback improved the quality and reliability of AI-generated outputs over time.
The research team sees potential for the framework to be adapted across a range of scientific domains, including medicine, public health, climate science, toxicology, and education research. They also emphasized the importance of transparency and human oversight in AI systems for science, given growing concerns around hallucinations, reproducibility, and automated misinformation in generative AI.
The broader vision behind MetaBeeAI is an approach to AI that works alongside human expertise rather than supplanting it. Rather than removing researchers from scientific decision-making, the system keeps them at the center of the process—using AI to handle scale while humans provide judgment, interpretation, and accountability.
That balance between automation and oversight may prove increasingly relevant as the volume of scientific literature continues to grow and as institutions consider how AI can responsibly support research, evidence-based policy, and knowledge creation across disciplines from environmental science to medicine.
MetaBeeAI combines large language models with human validation to help researchers automatically extract, organize, benchmark, and analyze information from hundreds or thousands of full-text scientific papers, while maintaining traceability and scientific rigor. The platform identifies relevant information within papers using AI, then allows experts to verify, correct, and refine outputs through a transparent review interface linked directly to the original source text.
"The system is not designed to replace scientists. Instead, MetaBeeAI aims to support researchers by reducing repetitive workloads and helping experts focus on interpretation, decision-making, and scientific insight," said Rachel Parkinson, lead author on the research published in Ecological Informatics.
The system was tested using nearly 1,000 scientific papers on pesticides and bees. It performed particularly well on factual extraction tasks, such as identifying bee species and pesticide compounds. Iterative expert feedback improved the quality and reliability of AI-generated outputs over time.
The research team sees potential for the framework to be adapted across a range of scientific domains, including medicine, public health, climate science, toxicology, and education research. They also emphasized the importance of transparency and human oversight in AI systems for science, given growing concerns around hallucinations, reproducibility, and automated misinformation in generative AI.
The broader vision behind MetaBeeAI is an approach to AI that works alongside human expertise rather than supplanting it. Rather than removing researchers from scientific decision-making, the system keeps them at the center of the process—using AI to handle scale while humans provide judgment, interpretation, and accountability.
That balance between automation and oversight may prove increasingly relevant as the volume of scientific literature continues to grow and as institutions consider how AI can responsibly support research, evidence-based policy, and knowledge creation across disciplines from environmental science to medicine.