Skip to main content
If you’ve saved hundreds or thousands of papers, Deep Search can analyze them as a corpus. This tutorial runs a population gap analysis across a ~400-paper Library. You’ll learn how to:
  1. Attach your whole Library to a search
  2. Run Deep Search scoped to your saved papers
  3. Follow up against the broader index to fill the gaps

​
Before you start

Open My Library and add papers by uploading, pasting DOIs, or importing from Zotero (which brings in full text too). Everything you upload is private and isn’t used to train models.

​
Step 1: Ask a Deep question of your Library

From inside My Library, the Library is already attached to the search bar. Type your question and toggle on Deep.
Deep Search samples your Library, finds an initial 20 relevant papers, then generates and runs its own sub-questions systematically across your papers. Every search is tagged with your Library, so it never touches the broader index. In the example it found 109 eligible papers and read 50. The answer includes a visual map of where studies are missing by population. In the example: no longitudinal data, few sex-specific analyses, and thin intervention data for race, ethnicity, and special clinical populations.

​
Step 2: Fill the gaps from the full index

Remove the Library from the search bar and ask:
Consensus keeps the gaps as context and searches all 220 million papers.
A regular Pro message works for small Collections (see Fill Collection gaps). Use Deep when the Library is hundreds or thousands of papers.

​
What Deep Search does behind the panel

The right-hand panel shows the plan as it runs: papers identified, screened, eligible, and included, then each phase (initial survey, clarified constructs, rephrased synonyms, contrasting evidence) with its searches and eligible-paper counts. Every search here is tagged with your Library.
The Deep step panel showing retrieved, eligible, and included counts and the searches it ran

A Deep Search plan panel: 494 identified, 467 screened, 412 eligible, with each phase's searches.

Deep is a longer, more robust version of a Pro search: up to 20 searches in parallel, multiple Research Agent tools like citation crawl, and 50 to 100 papers read. The answer’s shape follows your prompt: a synthesis, a comparison, a critique, a reading list, an evidence map, or a custom report.

​
Export the report

Scroll to the bottom and click Export: Copy text, Copy with citations (APA, MLA, Chicago, Harvard, LaTeX, AMA/Numeric), or Export as PDF. Set the in-line citation style (author-year or numeric) from the thread settings button at the top of the page.
The Export menu at the bottom of a Deep report

Deep Search

Library