Google Research has released Deep Review to the public. In a blog post dated 14 October 2026, research scientists Rajesh Jayaram and Vincent Cohen-Addad describe it as an agentic AI system that gives “rigorous feedback and technical verification on complex documents,” expanding from computer-science papers into domains such as legal and finance.
From conference pilots to a public tool
Deep Review grew out of Google’s Paper Assistant Tool (PAT). Over the past year, Google partnered with STOC 2026, ICML 2026, NeurIPS 2026 and ICLR 2027 to offer private, pre-submission feedback to authors of more than 16,000 manuscripts who opted in. Google says that feedback was “completely decoupled from the official conference review process,” and that PAT did not collect the data for training.
How to use it
Every user gets two complimentary document reviews per week, Google says. Heavier users can connect a Gemini API key for higher limits and use the Deep Review API. Inputs at launch are PDFs and plain text, with multi-file support “coming soon.”
On privacy, Google says documents uploaded to Deep Review “are never used to train, fine-tune, or improve Google models,” and are stored for one year only so they can be shown again in the Deep Review interface.
Why Google says it built it
Google argues that AI has sped up writing papers and other technical documents while human review capacity has not kept pace. As its own illustration, it says there were “over 74,000 submissions to three major AI conferences (ICLR, NeurIPS, and ICML)” this year, up from about 45,000 in 2025, 32,000 in 2024 and 23,000 in 2023. Those counts are Google’s figures.
Deep Review runs a four-stage pipeline: it segments a document, assigns more compute to denser sections, verifies segments in parallel, then synthesises the findings and grounds claims. Google also points to a related “Teamwork Deep Document Review Pattern” for Antigravity users.
What this means
- Authors can try it without a conference invite. The public release opens a tool that had been limited to opted-in pilot authors.
- It is a helper, not the official review. Google stresses the conference pilots were separate from formal peer review.
- Limits and privacy terms are stated up front. Two free reviews a week, optional Gemini billing for more, and a one-year storage policy with no training use, according to Google.
Sources
- Google Research (Rajesh Jayaram and Vincent Cohen-Addad), “Analyzing technical documents with Deep Review”, 14 October 2026. https://research.google/blog/analyzing-technical-documents-with-deep-review/
