A place for the evidence you read.
Start with the question
Define the review question and eligibility criteria before screening. Keep exact search strings, databases, dates, filters and export details in Search notes. This tool does not search databases, retrieve PDFs, verify claims or make inclusion decisions.
The built-in studies are clearly labelled fictional demonstrations. They are not published evidence and must not be cited. Replace them by creating a new review or importing your own references.
Import references
RIS supports common title, author, year, DOI, URL, journal, abstract and keyword tags, including continuation lines. BibTeX supports braced or quoted literal fields, nested braces, numeric fields and brace/parenthesis entry delimiters. Expand @string macros and concatenated strings before importing. TeX accent commands and other formatting are not fully rendered; inspect imported text.
CSV must be UTF-8 and comma-separated, with a title column. Quoted commas, escaped quotes, a BOM and multiline fields are supported. Common headers include title, authors, year, doi, url, journal, abstract, design, population, intervention, comparator, outcomes, findings, limitations, appraisal and notes. Authors use semicolons between names. Year must be empty or four digits.
File selection loads text into the import box; click Import references to add records. All imported references start unscreened and receive new local IDs. CSV/RIS imports do not restore screening decisions; use the JSON review backup for that. Unsupported or invalid records reject the batch without changing the open review. Up to 2,000 references per import and 3,000 stored records per review are supported. Reference files must be under 8 MB.
Review duplicate candidates
DOIs are normalised by removing resolver prefixes and whitespace and using lowercase. Candidate groups share a DOI or normalised title and year, including connected groups with both kinds of matches. Missing years can therefore produce candidates. Versions, corrections and distinct reports may be grouped together; review them yourself.
Choose which record to keep active, then mark the others as duplicates. No records or extraction notes are merged or deleted. Marked records remain stored, retain their original data and link to the selected active record. They are excluded from active screening counts and the evidence matrix. Restore active reactivates a record. Up to 100 candidate groups are displayed at once; resolving them reveals remaining groups.
Removing a reference from the library is a separate explicit action. Its linked duplicate records are restored to active status. Undo action restores the prior project snapshot. Download a backup before large changes.
Screen in two stages
Use Title / abstract screening first. Include, Exclude, Maybe and Unscreened are available. Exclusions need a reason. Only title-stage included records can receive a full-text decision. Full text is read outside this tool; source and DOI links are optional navigation aids, not fetched content.
Changing a title decision away from Include clears the current full-text decision and reason. Full-text Include counts as final included evidence in the matrix’s default scope and reports. Maybe is a temporary judgement, not a final inclusion. Current counts are a workspace snapshot, not a complete PRISMA flow, registration or review-methodology certification.
Extract and compare
The matrix has design, population, intervention/exposure, comparator, outcomes, findings, limitations, appraisal and notes fields. Choose final included, title-stage included or all active records as the scope. Up to 25 records appear per page. Evidence CSV exports the full selected matrix scope, not just the visible page.
Appraisal is your own free-text judgement. The app does not apply a validated risk-of-bias instrument, rate certainty, perform meta-analysis or infer causal conclusions. Record measurement details, effect estimates, uncertainty, follow-up and source page references where relevant to your review.
Backups and handoff
Save here stores up to ten reviews in this browser, subject to quota. Saving an existing review ID updates that saved copy. Download a JSON review backup to preserve references, screening, duplicate links, extracted fields and the action log. Open review restores the complete file, up to 16 MB. The app does not automatically save.
Reference CSV and RIS exports use your chosen export scope. RIS output uses generic reference records and preserves basic metadata rather than the original complete reference type/fields. The Markdown and printable reports contain the review plan, counts and final included evidence. Use the browser print dialog’s Save as PDF destination for a PDF.
The action CSV includes recorded imports, screening and structural changes and evidence-field edits. Only the latest 10,000 actions are retained. Typing review-plan notes is not logged separately. Undo action restores earlier snapshots, including their action log. This editable local project is not a tamper-proof audit trail or a collaborative review system.
Privacy and limits
There are no accounts, uploads, analytics, external fonts, remote libraries or background API calls. Reference and evidence data stay in the browser, except when you explicitly open external source links. Browser storage and downloaded files remain accessible to people using your device or profile. Keep backups and follow your organisation’s data-handling requirements.
This first release does not read or store PDF files, annotate full text, automatically format a citation style, sync collaborators or carry out literature searches. Use content you have permission to reproduce.