AI-Assisted Research into Alternative Life: An Evidence-Grounded Computational Framework for Oxygen-Independent Metabolism and Hypothetical Biochemistries

Background: Research on alternative life asks which features of familiar biology are universal and which are accidents of Earth's history. Artificial intelligence (AI) can speed up literature synthesis and hypothesis generation in this field, but language models can fabricate citations, omit caveats and propose chemically impossible reactions.

Problem: Public discussion often conflates two very different claims: that an organism does not require molecular oxygen (O₂), which is well established for anaerobes, and that an organism contains no oxygen atoms at all, which is speculative and undemonstrated.

Approach: This paper proposes a reproducible, human-supervised workflow combining source-linked evidence extraction, a four-level evidence taxonomy, constraint-based metabolic modelling with COBRApy, automated atom- and charge-balance checks, and an uncertainty-aware feasibility rubric. Known oxygen-independent metabolism serves as the benchmark domain.

Contribution: The contribution is a transparent method and evaluation design, not the discovery of a new organism. The framework keeps claims proportional to evidence by separating what is demonstrated, what is inferred and what is speculative.