Jev Reviewer

What does this paper actually report?

Drop a paper here, with its supplements.

Data extraction for systematic reviews and meta-analyses. Add the trial report with its appendix, protocol, analysis plan and data tables (PDF, Word, Excel, PowerPoint, OpenDocument, RTF, web pages, CSV or text), then ask for what your form needs. Every file is searched; each answer is a quote from one of them, with its page or row, for you to check.

Files are kept in this browser only. Nothing is uploaded.

The sample is Johnson et al. 2026, PLoS Medicine, with its statistical analysis plan and CONSORT checklist (CC BY 4.0).

How it works

Jev, TypeSafe's System One model, never writes an answer. It reads each file a few pages at a time and points at the lines that answer your question; the app copies them word for word, with their file and page. Each question takes two passes: every stretch of every file is screened for all questions at once, then each promising line is checked on its own, so an answer comes back Reported, Unclear or Not found, and one spread over several table rows keeps all of them.

You stay the reviewer. On each quote, the green tick makes it the question's answer (the other quotes fold away until you ask for them), the pencil opens your answer to edit, and the squares copy the quote with its place. The quotes stay word for word; your answer is what you edit. A question marked for outcome data is answered arm by arm (N, mean and SD, or events and N), filled from the numbers in the quotes. The Check button under a question checks it as it stands, for example when the files do not report it. The extraction table shows every study against every question, counts the PRISMA flow once studies are excluded, exports the results, and, in its tabs, shows risk of bias, retractions and open access copies, compares your answers with a second reviewer's, and drafts your methods paragraph, the PRISMA 2020 flow diagram and the table of included studies. Before screening, Deduplicate search results (in the column) removes the duplicates across your database exports that both the rules and Jev find, and lists those only one of them finds for you to decide. Screen titles and abstracts then puts each record against your eligibility criteria: Jev judges each criterion from the title and abstract, you decide every record, and the included ones become the project's studies.

Projects, studies, files and answers stay in this browser, on this device: nothing is uploaded or kept on a server. Only the text of the files and your questions go to TypeSafe, through this site's relay, which pays with a shared key up to a daily limit. To check for retractions and open access copies, a study's DOI or PubMed id goes to Crossref, OpenAlex and PubMed, and through the relay to Retraction Watch (XeraRetractionTracker) and PubMed Central. In Chrome, speech recognition sends your voice to Google. Google Analytics counts visits, and sees neither your files, your questions nor the names of your projects and studies.

Enter asks. Find looks up exact words in the files, at once and without Jev. Say a question, or “next” and “previous” to move between quotes.

Keys

When no field is being typed in: j and k move between answers, c checks or unchecks the one in view, e opens or closes its answer editor, n goes to the next answer not yet checked, and / goes to the question box. In the editor, Esc closes it. Once every answer is checked, the next study to check is one press away.

The guide walks through a whole review, from importing references to reporting how the tool was used.

Projects

A project holds its studies; a study, its files and answers. They are kept in this browser, on this device, and never uploaded. A questions file loaded in a project runs on any of its studies, and the project's export has every study's answers in one sheet.

Question templates

Ready-made questions for common extraction forms, in plain words Jev can search for. Added to , they join its list; questions already there stay once. Download one to edit it, then upload it.

Extraction table

Project

Risk of bias

Study

Each domain lists your answers to its questions from the risk of bias template (open one to check it). Judge each domain; pressing a judgment again clears it.

Deduplicate search results

Add the exports of your database searches (RIS, BibTeX, EndNote, PubMed, Web of Science, CSV or Excel). Duplicates are found two ways: by their DOI, PubMed id and title, and by Jev. A pair both find is removed; a pair only one finds is left for you to decide.

Title and abstract screening

Project

Eligibility criteria

Jev judges each criterion from the title and abstract: met, not met, or not said. A changed criterion is asked again; the others keep their answers.

Keys: j k move, i include, m maybe, x exclude.

    Settings

    Keys come from console.typesafe.ai.