Covidence is a web-based platform built specifically for systematic reviews that carries a team from imported search results through title and abstract screening, full-text review, risk-of-bias assessment, and data extraction in one place. Unlike a general reference manager, it is designed around the two-reviewer model: it shows each record to two people independently, hides their decisions from each other until both have voted, and surfaces the disagreements for resolution. For many review teams Covidence is the workbench where the screening and extraction stages actually happen.
The screening workflow Covidence is built around
Covidence organises a review into stages that mirror the methodological steps a review must follow, so the software enforces the order rather than leaving it to a spreadsheet. You import the deduplicated records, screen on title and abstract, promote the survivors to full-text review, record the reasons studies are excluded at full text, then move included studies into extraction and appraisal. Each transition is a gate, and the running counts at every gate feed straight into reporting. If you want the wider picture of where these tools sit, our overview of systematic review screening tools compares the category, and the broader sequence is laid out in our guide to the steps in a systematic review.
Title and abstract screening, then full text
The first pass in Covidence is fast and binary. Each reviewer sees a title and abstract and votes yes, no, or maybe, working through the queue without seeing the other reviewer’s call. Records both reviewers accept advance automatically; records both reject drop out; and anything the two disagree on is held for resolution. The same independent logic repeats at full-text review, except here Covidence requires a recorded reason for exclusion for every study screened out, which is exactly what the flow diagram later needs. The discipline of the two stages is worth understanding in its own right, and our guides to title and abstract screening and full-text screening explain what each pass is trying to achieve.
Dual independent screening and conflict resolution
The feature that distinguishes a screening platform from a reference manager is blinded dual screening. Covidence assigns two reviewers to each record and conceals their votes from one another until both have decided, which prevents one reviewer’s opinion from anchoring the other. When the two disagree, the conflict goes to a dedicated resolution view where a third person, or the original pair in discussion, settles it. This matters because inter-reviewer reliability is a reported quality signal; if you are documenting agreement you can take the decision data into a Cohen’s kappa calculator to quantify it. The number of reviewers a stage needs is itself a methodological choice, covered in our guide on how many reviewers to use for screening.
Extraction templates and risk-of-bias integration
Once studies are included, Covidence moves them into data extraction. You build an extraction template, the set of fields you want from every study, and two reviewers extract independently into that template before reconciling, so the same consensus logic that governs screening also governs the numbers that enter your synthesis. Covidence also supports risk-of-bias assessment alongside extraction, so each included study can carry its appraisal in the same record rather than in a separate file. The design of the template repays care, because a vague field invites inconsistent answers; our guide to data extraction in a systematic review explains how to specify fields so two reviewers record the same thing.
Reporting and export
Because Covidence tracks every count as records move between stages, it can generate the figures a PRISMA flow diagram requires, including the number imported, the duplicates removed, the records screened, the full texts assessed, and the studies excluded with their reasons. The platform exports the extracted dataset and the decisions for analysis elsewhere, and it produces a draft flow diagram you can refine. Capturing these numbers as you go, rather than reconstructing them at the end, is one of the practical reasons teams adopt a dedicated tool in the first place.
Pricing, access, and how it compares to Rayyan
Covidence is a paid product rather than a free one, and most users reach it through an institutional subscription: many universities and research organisations hold an account that covers their members, so an affiliated researcher can often start a review at no personal cost by signing in through their institution. Cochrane authors also have access arranged through Cochrane, since Covidence is a closely connected tool in that ecosystem. Individuals without an institutional licence can typically run a limited number of reviews before needing a paid plan, but pricing and allowances change over time, so check the current terms rather than relying on a remembered figure. The most common comparison is with Rayyan, which is free for its core features; many teams choose Covidence when they want guided end-to-end workflow and built-in extraction, and choose Rayyan when budget rules and they only need screening. Our guide to screening with Rayyan sets out that trade-off in detail.
Is Covidence an artificial intelligence tool?
Covidence is best understood as a structured workflow and collaboration platform rather than an automation engine. Its core value is process: enforcing two independent reviewers, hiding their votes, surfacing conflicts, and keeping counts. The platform has added machine-assisted features over time, and the wider screening field is moving toward relevance ranking and other automated aids, but the human decisions and the auditable two-reviewer record remain the point. Treating it as a rigorous bookkeeping and refereeing system, not a black box that decides for you, is the right mental model.