Use Guide

On your first visit, aim for one small, traceable research run. Start with one question, save the papers you plan to use, inspect the evidence at page level, and finish with a draft or deliverable that can be checked later.

1. StartOne research question
2. CollectA saved paper set
3. VerifyPage-level evidence
4. FinishA draft or deliverable

1. Choose the right entry point

If you enter a protected workflow before signing in, NNScholar sends you to the main website for authentication and then returns you to the workflow you were trying to use.

2. Complete one minimal research run

Create a workspace for one research question, then ask Chat or Side chat for a short plan. Use the sequence below to move from discovery to a reviewable result.

Literature search: how to find relevant papers

Open in the appLeft Chat or right rail / Literature

Use Chat for a quick first pass, then switch to the dedicated Literature workspace when the search needs structure.

1. Describe the disease, population, method, exposure, outcome, or paper type in the chat box on the left.
2. For formal screening, click Literature in the right sidebar. Choose PubMed, Google Scholar, Wanfang, Semantic Scholar, or arXiv, then set fields and filters.
3. Review the metadata, open the source, download available full text, and save selected papers or batches to Library.

Done when

Useful papers are saved to a named Library folder, and the query and filters can be repeated.

Open the detailed module guide

Paper reading

Open in the appRight rail / Paper reading

Open Paper reading once a result is relevant enough to inspect at page level.

1. Open a PDF from Library, upload a local file, or select a PDF artifact from the workspace.
2. Ask about one section, table, figure, endpoint, model, or claim at a time. Request page numbers when the answer may support writing.
3. Use PDF search, Quick Translate, selected-text translation, highlights, notes, and the annotation panel to capture what matters.

Done when

Each manuscript-ready finding has a page number, table or figure label, and a short evidence note.

Open the detailed module guide

Library

Open in the appRight rail / Library

Library is the working collection for papers you plan to read, compare, cite, or exclude.

1. Save records with their abstracts, authors, journals, source links, publication dates, and identifiers.
2. Organize by research decision. Folders such as methods, screened-in, excluded, or figure-sources are easier to act on than broad topic folders.
3. Upload local files and wait for the indexed state before attaching them to grounded chat.

Done when

Core references, exclusions, missing PDFs, and next-read papers are visibly separated.

Open the detailed module guide

Document Writing

Open in the appRight rail / Document Writing

Move here when the output needs to become a manuscript, response letter, cover letter, or other formal document.

1. Set the target journal and document type. Bring in the relevant papers, PDF notes, analysis outputs, tables, figures, and Deliverables.
2. For a new manuscript, establish the structure before drafting sections. For an existing manuscript, import the DOCX for revision or format matching.
3. Use document Q&A and separate review passes for logic, unsupported claims, inconsistent results, references, and formatting.

Done when

The document has a defined target, reviewed structure, and source trail for every substantive claim.

Open the detailed module guide

Research experts

Open in the appRight rail / Expert library

Use a research expert when the task needs a defined source corpus or a repeatable specialist workflow.

1. Choose an existing expert, or create one from a Library folder, a PDF set, or other source files.
2. Check the source list, field, intended use, and limits before relying on the expert.
3. Bring it into Chat for search strategy, evidence extraction, methods critique, citation checks, manuscript revision, or reviewer-response planning.

Done when

You know which source corpus supports the expert and which parts of its output still need review.

Open the detailed module guide

Automated literature recommendations

Open in the appRight rail / Research automation

Set up Research automation when the same literature search needs to run again.

1. Define the topic, source query, time window, schedule, output language, and review destination.
2. Use a Library folder as seed context when the recommendation should stay close to an existing paper set.
3. Review each digest, save useful papers, and adjust the query when the result set becomes noisy.

Done when

The query, schedule, destination, reviewer, and next action for useful papers are all defined.

Open the detailed module guide

Browser automation

Open in the appRight rail / Browser

Use the embedded Browser for publisher pages, DOI records, databases, guidelines, software documentation, and journal portals.

1. Open the page from the right rail and inspect the source inside the research workspace.
2. Save paper records to Literature or Library, and move requirements or evidence into Document Writing or Deliverables.
3. For submission, use the Automatic submission assistant to prepare the checklist, then work through the journal portal in Browser.

Done when

Web evidence, portal requirements, and the submission checklist are saved back to the workspace.

Open the detailed module guide

Example Demo: First Complete Run

Suppose you want to study whether PCAT on coronary CTA predicts cardiovascular events in patients with diabetes. Run a minimal project like this:

1. Create a desktop workspace named diabetes-pcat-prognosis.
2. In Chat, enter the research idea and ask for a search, reading, study design, analysis, and writing plan.
3. Open Literature to find related papers, save core references, and download PDFs.
4. In Library, create core, methods, and background folders.
5. Use Paper reading to open core PDFs and ask for main results, statistical models, and citable page numbers.
6. Return to Chat and use phase skills to generate the study design and analysis plan.
7. In Document Writing, draft or revise the manuscript, match formatting, ask document questions, and run document review.
8. Use the Automatic submission assistant + Browser to check the journal portal and save the checklist into Deliverables.

The point of the demo is that every step leaves a reviewable file, evidence item, or manuscript version instead of only a chat answer.

3. Plan before agent execution

Planning is not overhead. It moves the unavoidable review work earlier: first inspect the structure, then let the agent execute.

When the task has more than one step, ask:

Make a short research plan before executing.
Keep it source-grounded.
List the papers, PDFs, or files that need to be checked.
Do not draft final claims until the evidence is reviewed.

If the plan changes repeatedly, summarize what you learned and start a new Side chat for a cleaner second pass. A clean re-plan usually beats patching a confused thread.

4. Use the right rail

The desktop right rail is the research workflow switcher:

• Chat: plan, compare sources, summarize progress, and decide next actions.
• Expert library: run specialized academic workflows for search strategy, paper reading, review design, or manuscript support.
• Literature: search literature, screen candidate papers, and create reading sets.
• Library: manage saved papers, collections, PDFs, and source links.
• Paper reading: read deeply, translate, annotate, and ask PDF-specific questions.
• Document Writing: draft and revise manuscripts, match formatting, ask document questions, and review manuscript quality.
• Deliverables: collect manuscript versions, figures, evidence tables, submission materials, and agent-generated files.
• Research automation: schedule alerts, reminders, and recurring reports.
• Plugins and skills: manage plugins, apps, skills, and external capabilities.
• Browser: inspect publisher pages, databases, and external sources.
• Side chat: use a workspace snapshot for focused reasoning or re-planning.

5. When to use skills

Skills are best for repeatable, reviewable research tasks with a stable procedure: literature search strategy, paper reading, evidence-table extraction, manuscript polishing, citation checking, or reusable workflow generation.

Do not treat a skill as a black box. Before using one, check what sources it reads, what it produces, and whether the output needs human review. For high-stakes reviews and manuscript work, return to the original papers before relying on the result.

Common paths

• Quick search: WebUse or Literature -> save papers -> Library -> Paper reading.
• Deep reading: Paper reading -> annotate and ask PDF questions -> Chat -> Deliverables.
• Evidence review: Library -> compare claims in Chat -> verify citations -> Deliverables.
• Manuscript writing: PDF evidence and study design -> Document Writing -> Deliverables.
• Specialized workflow: Expert library -> choose a research skill -> run the workflow -> review outputs.

Completion checklist

• The research question and deliverable are explicit.
• Important claims point back to saved papers, PDF evidence, manuscript versions, or deliverables.
• Search terms, inclusion choices, and exclusion reasons are recorded.
• Agent output has been checked against sources.
• Final files, tables, manuscript versions, or submission materials are visible in Deliverables.
• Recurring work is captured in Research automation.

Before you upgrade

Basic literature search can start free. Heavier AI workflows, subscriptions, batch actions, advanced research-agent usage, and some skill or plugin workflows consume credits. Check the pricing page and account settings before relying on a paid workflow for production research.