Can AI Conduct Research: A Pragmatic Experiment

This study evaluated the capability of three prominent Large Language Model (LLM) AI tools—ChatGPT, Copilot, and Claude—to independently conduct a complete research process. The investigation involved generating hypothetical data (“silicon samples”), performing inductive thematic analysis, and composing research reports. Findings reveal that while AI can produce outputs resembling research products, the quality varies significantly, often lacking depth, accuracy, and synthesis. Notably, AI-generated data and analyses tend to be predictable, with issues such as hallucinated references and misquoted data, underscoring the necessity of human oversight. The study highlights both the potential and current limitations of AI in autonomous research, emphasizing that human researchers remain essential for producing high-quality, impactful scholarly work.