A Survey of AI Scientists

TL;DR

The paper addresses fragmentation in the AI Scientist field and proposes a unified six-stage framework to model end-to-end autonomous research. It analyzes 50+ systems across 2022–2025, revealing patterns of stage coverage and a three-phase historical progression from foundational modules to scalable collaboration. By unifying methodology across six stages and providing a comprehensive taxonomy and evaluation lens, the work clarifies current capabilities and gaps, and outlines open problems in reproducibility, uncertainty, cross-domain generalization, and governance. The resulting roadmap highlights how end-to-end autonomous scientific systems can become trustworthy, collaborative partners that accelerate discovery while preserving scientific integrity.

Abstract

Artificial intelligence is undergoing a profound transition from a computational instrument to an autonomous originator of scientific knowledge. This emerging paradigm, the AI scientist, is architected to emulate the complete scientific workflow-from initial hypothesis generation to the final synthesis of publishable findings-thereby promising to fundamentally reshape the pace and scale of discovery. However, the rapid and unstructured proliferation of these systems has created a fragmented research landscape, obscuring overarching methodological principles and developmental trends. This survey provides a systematic and comprehensive synthesis of this domain by introducing a unified, six-stage methodological framework that deconstructs the end-to-end scientific process into: Literature Review, Idea Generation, Experimental Preparation, Experimental Execution, Scientific Writing, and Paper Generation. Through this analytical lens, we chart the field's evolution from early Foundational Modules (2022-2023) to integrated Closed-Loop Systems (2024), and finally to the current frontier of Scalability, Impact, and Human-AI Collaboration (2025-present). By rigorously synthesizing these developments, this survey not only clarifies the current state of autonomous science but also provides a critical roadmap for overcoming remaining challenges in robustness and governance, ultimately guiding the next generation of systems toward becoming trustworthy and indispensable partners in human scientific inquiry.

Figures (9)

• Figure : The Architectural Landscape of the AI Scientists for Automatic Research. The main 4x6 matrix maps the six methodological stages of the scientific workflow (horizontal axis) against four top-down layers of abstraction, from Applications to Models (vertical axis). The panel at the bottom illustrates the field's three-phase historical evolution, categorizing representative works to provide a chronological perspective on the development of AI Scientist systems.
• Figure : Evolution of AI Scientist research (2022–2025). A horizontal timeline illustrates three major phases: (I) Foundational Modules (2022–2023), (II) Closed-Loop Integration (2024), and (III) The Frontier: Scalability, Impact, and Collaboration (2025–present). Each phase highlights representative systems, with upward arrows denoting increasing levels of autonomy and integration.
• Figure : End-to-end workflow of an AI Scientist system. The six stages represent a closed scientific loop, starting from knowledge synthesis and ending with validated scientific reports. Arrows denote data and reasoning flow, while the outer frame indicates embedded reflection and evaluation mechanisms.
• Figure : Pipeline for Automated Literature Review. The workflow transforms unstructured scientific corpora into structured knowledge through five sequential stages with feedback mechanisms enabling iterative refinement.
• Figure : Pipeline for Idea Generation. The workflow transforms structured knowledge into testable hypotheses through four sequential stages supporting creative discovery and feasibility evaluation.