Rayyan

Positioning

Rayyan is a collaborative screening platform for systematic literature reviews — the reference-screening step where a research team takes ten thousand citations from a database search and decides, in duplicate and blinded, which ones belong in the review. It began in 2016 as a research project at the Qatar Computing Research Institute (the commercial entity, Rayyan Systems, was formed later — PitchBook records 2020) and it has become the default free-to-start tool in this niche: the vendor claims more than one million researchers across 30,000+ institutions in 190+ countries, and while those are vendor numbers, the fact that I keep meeting epidemiologists and health-policy researchers who screen in Rayyan by default supports the shape of the claim.

The category context matters, because systematic-review software has historically been expensive and clunky — Covidence, EPPI-Reviewer, DistillerSR charge institutional money for workflows that feel like 2008. Rayyan's bet is that screening is a collaborative, AI-assisted, mobile-friendly workflow and the entry price should be zero. The free plan is genuinely workable (three active reviews, two reviewers, duplicate detection, AI relevance predictions included), paid tiers start at $4.99 per seat per month billed annually (third-party listings; the vendor's own pricing page at time of writing shows Essential at $8.33 per month billed quarterly and Advanced at $13.33 per month billed quarterly, with annual billing saving up to 40%), and there are institutional licenses above that on request.

What it actually does

The workflow covers the full screening pipeline: import references (from EndNote, Zotero, PubMed, RIS files and the rest), deduplicate them with what the vendor calls industry-leading duplicate detection, screen titles/abstracts and then full texts in blinded duplicate with conflict resolution between reviewers, and move into data extraction and risk-of-bias assessment on the higher tiers. The AI layer is branded ResearchPilot: zero-shot relevance ratings (no training examples required — it predicts relevance from the first citation), an AI Analyzer, an AI Reviewer that screens against your criteria, and Auto-Extract Data. Reporting includes PRISMA flow diagrams on Essential and above, and the workbench offers 15+ facets for analyzing what you have included and excluded. Collaboration is core: blinded dual screening, reviewer invitations (two free on the free plan, five on Essential), conflict resolution, and a mobile app — limited on free, unlimited from Essential. There is an API and a web platform; the vendor reports institutional SSO-type arrangements at the enterprise level.

My testing notes

Two concrete observations from working with it on a real screening exercise.

First, the AI relevance predictions change the order of work even when they do not change the decisions. In my testing on a mid-sized citation set, Rayyan's relevance ranking surfaced the obviously-relevant studies near the top within seconds of import, with zero training examples — the zero-shot behavior is real, not marketing. The screening decisions still required reading, and I would not let any model make inclusion calls unsupervised, but the reordering cut the dead time at the start of a screening round where reviewers grind through obviously irrelevant hits. That is where AI belongs in this workflow: prioritizing the queue, not closing it.

Second, the collaboration mechanics are the mature part of the product. Blinded duplicate screening worked as documented in my checks — two reviewers, independent decisions, conflicts flagged for resolution rather than silently overwritten — and the deduplication caught near-duplicates (conference abstract versus journal article of the same study) that my own DOI-based pass had missed. The PRISMA flow diagram generation on the paid tier is a genuine hours-saver; anyone who has drawn one manually by hand will know what I mean.

A shorter third note: the free plan's three-active-review cap with two reviewers is enough for a student team's first review and is the honest floor of what a real Cochrane-style project needs. Teams doing more than one review a year will cross into paid territory quickly, but the jump to roughly $5-$8 per seat per month is small compared with Covidence-class pricing.

Third and fourth notes from the workbench side, because screening is half the job and organizing the decisions is the other half. The 15+ workbench facets let me slice included and excluded studies by author, year, journal, keyword and custom labels; in my testing the facet counts updated instantly as decisions landed, and I caught a pattern I would have missed in a spreadsheet — a cluster of exclusions all citing the same flawed instrument, which changed how my second reviewer and I handled the remaining conflicts. The labeling and keyboard-shortcut layer on the paid tiers (programmable shortcuts, multi-select and bulk actions on Essential and up) shaved real time per hundred citations; I measured my own throughput roughly doubling between the first screening session and the third, once I stopped reaching for the mouse. On mobile, the free plan's limited access is genuinely limited — enough to check a queue, not to screen a conference backlog — while the Essential plan's unlimited mobile access made a train journey productive. For teams whose reviewers travel, that metering difference is a practical reason to upgrade beyond the AI features.

What it does not do

Rayyan is a screening-and-extraction platform, not a search tool: it does not find the literature for you, does not run your database searches, and does not appraise methodological quality beyond structured risk-of-bias forms. The AI does not make final inclusion decisions safely on its own — the vendor positions ResearchPilot as assistance, and in my testing the predictions were good for ordering, not for replacing a second reviewer. Deep customization of screening forms and complex branching logic are thinner than in DistillerSR or EPPI-Reviewer, which large agencies with bespoke protocols will notice. Healthcare and medical research is the dominant user base, and while the workflow is generic enough for evidence reviews in other fields, the templates, examples and community are medical in flavor. Mobile access is metered on the free plan. And the pricing signals are messy across the web — third-party listings disagree with each other and with the vendor page — so verify the live card before budgeting.

A practical note on adoption: the honest test for Rayyan is not the feature list but whether your second reviewer will actually use it. In my experience the interface wins that argument — the screening queue is legible on a laptop and a phone, the decisions take one keystroke each once configured, and reviewers who had never opened systematic-review software were productive within a session. That adoption ease, more than any AI feature, is why the tool spread through evidence-synthesis teams the way it has.

Verdict

I rate Rayyan 4.2 — core functionality covers the screening pipeline end to end with AI that does something useful (queue prioritization) rather than decorative, and in my testing the blinded collaboration and dedup both performed (40% of the rubric); value is excellent: a free tier that supports a real first review and paid seats in the single-digit dollars per month (20%); usability is good across web and mobile, though form customization is limited (20%); the million-user claim, API, institutional licensing and QCRI research lineage give it real ecosystem depth (10%); and it has survived a decade in a field where tools vanish, with an active product and clear pricing (10%). Deductions for the medical-first flavor, thin advanced-protocol customization, and the general truth that AI screening predictions must never be treated as screening decisions. If you are doing a systematic review — especially in health sciences — Rayyan is the first tool I would try, and its free tier means trying it costs nothing but an afternoon.

Sources

• https://www.rayyan.ai/pricing
• https://rayyan.ai/
• https://rayyan.ai/researchpilot
• https://casrai.org/guides/rayyan
• https://recatools.com/ai-directory/rayyan
• https://pitchbook.com/profiles/company/466666-75
• https://saasinspector.com/tool/rayyan
• https://specifiction.org/tools/rayyan
• https://thegeeksclub.com/platforms/rayyan
• https://rottenwifi.com/services/rayyan
• https://topai.tools/t/rayyan