Google Research published findings on October 7, 2026, from a three-month randomized controlled trial examining how generative AI assistance affects patent drafting and professional skill development among 133 practicing lawyers.
Short-Term Gains in Patent Drafting
Researchers David Autor and Tanya Rodchenko detailed a field experiment that randomized access to an unreleased Google Labs AI patent writing assistant tool—now part of Gemini Notebook—among 133 lawyers at 11 intellectual property law firms. Two-thirds of the lawyers received early access to the tool for 90 days, while the control group was withheld from using it during the study period.
After 10 days of AI assistance, participating lawyers drafted a hypothetical patent using inventor materials, scoring 0.34 standard deviations higher than the control group. By 90 days, this improvement increased to 0.38 standard deviations, driven by a lower frequency of poor scores and a higher frequency of good scores across quality dimensions.
Divergent Skill Building for Juniors and Seniors
To measure unassisted skill development after 90 days, researchers instructed lawyers to complete a redlining task where AI tools were prohibited, requiring them to correct errors in an existing hypothetical patent. Treated lawyers outperformed controls by 0.32 standard deviations on this unassisted task.
However, the unassisted performance boost was driven entirely by senior lawyers with seven or more years of experience, who outperformed control subjects by 0.45 standard deviations. Junior lawyers showed no average skill gain on the redlining task, with their scores splitting into more strong scores and more low scores.
Key facts and where they come from
- The study involved 133 lawyers across 11 intellectual property law firms.
among 133 lawyers at eleven intellectual property law firms
- Senior lawyers given AI tool access outperformed control subjects by 0.45 standard deviations on the unassisted redlining task.
who outperformed control subjects by 0.45 SD
- Junior lawyers showed no average skill gain, with their scores splitting into more strong scores and more low scores.
whereas junior lawyers show no average skill gain, their scores splitting into more strong scores and more low scores.