Learning with AI and web search
Authors: David Autor, Tanya Rodchenko, Josh Martin, Zanna Iscenko, Scott Strand, David Pearl, and Melissa Ferere.
Year: 2026 · Series: NBER Working Paper 35720 · DOI: 10.3386/w35720
Study: Pre-registered, three-month randomized controlled trial; 133 practicing patent lawyers at eleven U.S. firms. AI-assisted drafting and subsequent unaided judgment assessed by blinded expert patent attorneys. Funded by Google.
Findings: AI access improved drafting quality, with larger gains among junior lawyers. Subsequent unaided judgment improved among senior lawyers. Junior lawyers showed no average gain, with more high and low scores.
Abstract excerpt
Foundational expertise may be a prerequisite for extracting durable skill from AI-assisted practice.
This is another spot where it feels like AI assistance is very similar in kind to web search tools themselves.
HT: Atoosa Kasirzadeh.
See also: Matt Beane (2019), Shadow Learning: Building Robotic Surgical Skill When Approved Means Fail, Administrative Science Quarterly, 64(1), 87–123. My dissertation discussion.