Large language models combined with open-source algorithms for identifying prior art for obviousness analysis

The literature shows that patents often slow innovation, and patent parasites that patent commons innovation threaten the growth of entire industries, including additive manufacturing (AM) or 3D printing. AM has expanded through open-source hardware, software and shared material recipes, yet this cumulative innovation ecosystem is increasingly exposed to broad patents on AM materials. This article examines how obviousness doctrine should respond to algorithmic and large language model (LLM)-assisted prior-art generation in cumulative-innovation fields. Twenty-five U.S. AM materials patents and applications from 2016 to 2026 are mapped against an open-source algorithm for identifying obvious AM material prior art. The results suggest that many claims can be characterized as known printable material classes, predictable combinations of known feedstocks, routine additive use, or ordinary processing adaptations. The article argues that LLMs substantially reduces the difficulty of locating, parsing and applying prior art, providing a clear path for challenging obvious claims. A hybrid strategy is proposed as the strongest defense: deterministic, open-source databases for legally robust disclosure, combined with LLM-assisted claim parsing, semantic search and prioritization. Such systems could help patent examiners, researchers and open-source communities distinguish genuine inventions from predictable formulation work and preserve open access to the commons in many fields.