SLR Copilot organizes a review into setup, import, screening, full text, extraction, risk of bias and export. These workspaces help record work; completing a phase does not certify that the review's methods or conclusions are sound.

This guide explains the decisions to inspect at each stage. The methods overview explains why review planning and software controls are separate responsibilities.

1. Set up the review and team

Record the question, eligibility criteria and reviewer arrangements before using suggestions to make decisions. Confirm who can review, resolve conflicts and change the setup. The available settings must fit the team's protocol; a saved setup is not protocol registration or approval.

2. Import references and inspect the result

The import workspace supports citation-file import, source searches and DOI/PMID lookup. Check the preview and resulting record counts, unresolved identifiers and duplicate handling. Retain the original export and search record so that corrections remain traceable.

Importing references from another tool does not establish that votes, PDFs, extraction tables, reviewer identities or audit history transferred with them. Before switching an active review, test a permitted small export and reconcile the fields your team needs. Do not resume at an advanced phase unless the required state has actually been checked.

3. Review title and abstract decisions

The queue includes reviewer votes, AI suggestions and conflict-resolution controls. Bulk voting and bulk application of suggestions are also available. Inspect the selected scope and reasons before using either; a reviewer clicking a bulk action is not independent assessment of every record.

AI confidence does not determine an acceptable review method. Keep uncertain cases and missing information visible, and follow the team's agreed selection process. See screening decisions and source checks.

4. Obtain and assess the full text

The full-text workspace combines a PDF viewer with eligibility controls. Automatic retrieval is constrained by open-access availability; it does not promise access to every report. Inspect retrieval failures and use permitted institutional access or manual upload where appropriate.

Cochrane chapter 4 calls for independent final full-text eligibility assessment by at least two people for intervention reviews. Configure the work to meet the planned methods; software suggestions do not supply the second independent reviewer.

5. Extract and check the required data

Build fields for the review's data needs. The interface supports AI suggestions, reviewer acceptance and source quotes/page references when supplied, plus dual-extraction and consensus controls. Inspect each value in context and resolve discrepancies; the existence of a source link does not establish that the value or interpretation is correct.

Use Cochrane's data-collection guidance when planning extraction checks. Do not turn an unreported value into an invented answer.

6. Assess risk of bias with the supported instrument

The current in-app form is RoB 2. Do not plan on using other assessment instruments in this workspace. RoB 2 addresses a specific result from a randomized trial; it is not a universal quality score. Decide whether it fits the included designs and results before using this workspace.

7. Reconcile the export with the review

Inspect the review data, references and PRISMA reporting outputs against the underlying decisions. Check the version actually exported. A completed export does not establish that a review is ready for publication.

PRISMA 2020 is reporting guidance, not certification of conduct or quality. Describe the automation, reviewer checks and deviations that actually occurred. Do not attribute procedures or validation to the review simply because a tool offers them.

Before choosing or switching software

List the required review controls and transfer fields, then inspect whether the current workflow meets them. The documented tool comparison can help frame that evaluation. Test the fields and controls your review needs before committing to a transfer.

SLR Copilot is separate from INRA's narrative-review drafting. Its screening, extraction and assessment controls do not establish the quality of an NLR draft.

Updated October 19, 2026 to clarify product limits and source checks. Qualified methods review of this update is pending.

About the author

Marwan Taman is the founder and CEO of INRA.AI. Original author attribution is retained; it does not indicate methods approval of this update.