Patent examiners have had access to AI-driven prior art searching since late 2022, but its use is expanding and is now even required in some technology groups (art units). Chief Information Officer of the USPTO, Jamie Holcombe, touted 30% of references result from using SimSearch.

SimSearch, the AI-based Similarity Search tool available through the USPTO’s Patents End-to-End (PE2E) Search, generates prior art results using keywords or phrases informed by an applicant’s specification.

Despite the promise of better prior art searches, some examiners question the quality of the prior art references that SimSearch finds. Prior art searching for patent applications is notoriously difficult due to disparate, diverse descriptions even for similar subject matter, vagaries in language, and frequent gaps in technical descriptions.

Because SimSearch is not publicly accessible, its effectiveness in overcoming these challenges remains unclear. In particular, it is uncertain how often SimSearch produces primary prior art references used to support novelty or obviousness rejections. Nonetheless, ongoing advancements in AI models suggest that the effectiveness of SimSearch, or a tool like it, may improve as well.