Paper2Agent: Nature introduces system that transforms scientific articles into 'virtual authors'

The scientific journal Nature has announced an artificial intelligence system called Paper2Agent, which promises to transform "every article" into an active agent. This tool, presented as a "virtual corresponding author," is capable of answering researchers' questions and applying the methods described in the scientific work to new data.

The system, developed by Jiacheng Miao, James Zou, and colleagues, is described in a research article published in Nature on September 16, 2026. The goal is to overcome the limitations of traditional PDF files, making scientific articles more interactive and functional by integrating text, code, and data.

The basic idea is to make scientific articles, which today are often static like PDF files, much more lively and interactive. The Paper2Agent system represents an attempt to integrate the text, code, and data of a publication, turning it into an active entity. This "virtual author" not only answers questions about the content but can also collaborate with agents derived from other articles.

The research, published in Nature on September 16, 2026, is the result of work by a team led by Jiacheng Miao and James Zou. Although Nature's presentation of the news emphasizes the ability to transform "any article," the researchers themselves acknowledge that the experiments conducted go beyond what has been demonstrated so far, pointing to a plausible path for the future of scientific communication.