Generative AI has accelerated writing and experimentation across technical fields, but human review capacity has not kept pace. Reviewing long scientific manuscripts or complex legal agreements requires specialized expertise, creating a validation bottleneck where content is generated faster than it can be rigorously checked.
Google has announced the public release of Deep Review, an agentic AI system designed to provide rigorous feedback and technical verification on complex documents. Initially built for computer science academic paper analysis, it now supports documents across domains such as legal and finance. It can serve as an iterative tool during drafting and as a support for verifying claims and evaluating work during review.
Deep Review evolved from the Paper Assistant Tool (PAT), an experimental initiative that helped researchers stress-test manuscripts before submission. Over the past year, PAT partnered with four flagship computer science conferences, STOC 2026, ICML 2026, NeurIPS 2026, and ICLR 2027, to provide private pre-submission feedback to authors of more than 16,000 manuscripts who opted in. The pilots responded to a surge in submissions: 74,000 to three major AI conferences this year, up from about 45,000 in 2025.
The system checks logical consistency and mathematical statements, suggests improvements, grounds facts, and compares work with existing research. It handles the most cognitively demanding parts of review while leaving higher-level judgments to human experts. Every user receives two free document reviews per week, and high-volume users can connect a Gemini API key for higher limits and programmatic access. Initially, Deep Review supports PDF and plain text inputs. Google states that uploaded documents are never used to train its models and are stored for one year solely for display in the Deep Review interface.
Source: Google AI Blog · Summarized by HeadlinesBriefing