Deep review is a powerful research agent that conducts full literature reviews across 220 million academic papers in just minutes. It helps you critically explore a topic, identify major findings, surface counterpoints, and get a structured understanding of the field.
How does Deep work?
It works just how a researcher approaches a topic! Deep breaks down your question into multiple sub-questions, runs up to 20 unique targeted searches, and synthesizes a detailed literature review of the top findings. Here's what happens behind the scenes:
Breaks down your question into multiple sub-questions
Runs up to 20 targeted searches related to your topic
Reviews over 1,000 papers to select the most relevant ones
Synthesizes an in-depth analysis based on the most relevant papers for your query. Deep typically analyzes up to 50 papers, and may analyze up to 100 when a larger set of papers would improve the response.
Deep is designed to adapt to your prompt. For general research questions, it will often produce a structured literature review with sections like introduction, methods, results, and discussion. For more specific prompts, Deep may adjust the format to better match what you asked for, such as a comparison, critique, reading list, evidence map, or custom report.
It also includes interactive visuals that:
Show the consensus among the research
Highlight key authors
Extract key claims and their supporting evidence
Reveal potential research gaps
How to run a Deep review
Start a new search here
Toggle on Deep mode within the search bar
Enter your question and submit your search
Wait a few minutes while Deep conducts and synthesizes results
How to customize in-line citation format
Click the Thread setting button at the top of the page
Under Citation format, choose your preferred format:
Author and year
Numeric
To set your export citation style, click Default export format and select your preferred format.
How to copy or export the Deep report
Use the Export button to download or copy/paste your analysis into Word, Docs, Notion, or Obsidian, with tables, charts, citations, and headings all preserved.
Submit a query with Deep
Scroll to the bottom of the report
Click Export
Select 'Copy text', 'Copy with citations', or 'Export as PDF' depending on your needs
Note: When selecting 'Copy with citations', you can choose from APA, MLA, Chicago, Harvard, LaTeX, or AMA/Numeric formats.
How to export the search results
Run a search
At the top of the results on the right, click the downwards arrow (Export) button
Choose your export format:
CSV
RIS
Bibliography (APA, MLA, Chicago, Harvard, BibTeX, or AMA/Numeric)
Example Searches
Deep Availability
Free Plan: up to 3 Deep reviews per month
Pro subscribers: 15 Deep Searches per month
Deep subscribers: 200 Deep Searches per month
You can always check your usage and when it will reset form your profile > Settings > Subscription.
FAQs
What does the checkmark icon within the in-line citation mean?
This means that the full text of the paper was used for our analysis. No checkmark means the abstract was used.
I'm on the free plan. Why did my Pro message count go down when I used Deep?
For free users, using Deep also uses one of your Pro message credits. This is because Deep is an upgraded Pro message that runs a longer, more in-depth review. Free users receive 15 Pro messages per month and up to 3 Deep Reviews per month.
Why does my Deep report cite fewer papers than it analyzed?
Deep may analyze more papers than it directly cites in the final report. It uses the most relevant papers to build the analysis, but only cites the sources that best support the specific claims, findings, and comparisons included in the response.
For example, Deep may review up to 50 papers for a query, but cite fewer than 50 if only some are directly relevant to the final analysis. For some queries, Deep may analyze up to 100 papers when a larger set of papers would improve the response. However, the exact number depends on the query, the available research, and what is most useful for the analysis.
Still need assistance? Contact us at [email protected] or via the Support Chat.