Researchers have developed aiXiv, the world’s first open-access preprint platform specifically designed for research generated by AI scientists and robot scientists, complete with integrated AI peer review and revision cycles. The platform aims to revolutionize scientific publishing by addressing the growing challenges posed by the rise of autonomous AI-driven research. Traditional academic publishing was built for human authors and manual peer review—processes that are now struggling to keep pace with the volume and pace of AI-generated scientific work. Key bottlenecks include insufficient review capacity, lack of quality control in preprint repositories like arXiv, unclear attribution rules for AI contributions, absence of structured environments for early-stage research proposals, risks of prompt injection and bias in AI review, and the lack of a unified, scalable infrastructure for human-AI and AI-AI collaboration. aiXiv tackles these issues with a comprehensive, multi-layered system. It supports the submission, review, revision, and publication of both research proposals and full papers, all within a single, open, and interoperable platform. The system features a structured, multi-stage review process that enables continuous improvement of scientific outputs. It uses versioning and traceable quality trajectories to ensure transparency and accountability in the evolution of research. To enhance reliability, aiXiv employs a multi-model voting mechanism to reduce individual model bias and improve consensus. It incorporates retrieval-augmented generation (RAG) to align AI reviewer feedback with real scientific literature, ensuring relevance and accuracy. A layered prompt injection detection and defense system safeguards the integrity of the review process against manipulation. Experiments show that large language models can achieve up to 81% accuracy in pairwise paper review tasks—demonstrating their capacity not just to generate research but to critically evaluate it. This capability enables AI-generated work to undergo iterative refinement, significantly improving quality. In testing, 100% of revised papers and 80% of revised proposals showed measurable improvement, with acceptance rates rising from 10% to 70% for papers and from 0% to 45.2% for proposals. The platform is designed to support a collaborative ecosystem where human researchers, AI agents, and robot scientists can interact through standardized interfaces. It includes mechanisms for mixed human-AI review, ensuring diverse perspectives and strengthening trust. This hybrid model addresses concerns about the authenticity and credibility of AI-generated science, offering a path toward a more robust and trustworthy scientific record. The implications of aiXiv extend beyond publication. It represents a potential shift in scientific discovery itself. With AI scientists capable of rapidly generating and testing thousands of hypotheses across disciplines, the nature of scientific innovation may evolve from incremental progress to high-impact, breakthrough-driven exploration. The platform enables rapid iteration on ideas, potentially allowing AI systems to explore nearly all conceivable research directions—something human scientists cannot do at scale. Looking ahead, the team plans to enhance aiXiv with reinforcement learning to create a dynamic research environment where AI agents can learn from experience, adapt strategies, and develop long-term planning capabilities. Over time, the platform could become a vast repository of scientific knowledge, fueling the next generation of AI scientists with real-world feedback and historical context. The researchers also aim to extend the system into physical sciences. By integrating with automated lab systems, aiXiv could enable end-to-end autonomous experimentation in biology, chemistry, and materials science—where AI generates hypotheses, designs experiments, controls instruments, analyzes data, and writes papers, all in a closed-loop cycle. Beyond science, the team is exploring applications in business and social sciences, particularly in assessing whether AI scientists can autonomously identify, collect, and analyze relevant data for high-quality research. To ensure long-term sustainability and governance, the team is establishing a non-profit organization, aiXiv Organization, to oversee the platform’s development and uphold scientific integrity. They are inviting leading academics to join an advisory board and are seeking partnerships with institutions, foundations, and industry to support the platform’s growth. One of the paper’s co-authors is also launching a startup focused on commercializing AI scientist technology across industries, accelerating innovation through autonomous research workflows. The development of aiXiv marks a pivotal moment in the evolution of science. As AI systems begin to generate research at unprecedented scale and speed, platforms like aiXiv provide the infrastructure needed to manage, validate, and advance this new era of discovery—ushering in a future where science is no longer limited by human capacity, but empowered by intelligent collaboration between humans and machines.