Compare / Covidence

Using Covidence?

Using SyRev with Covidence

Covidence is a not-for-profit platform covering the whole systematic review workflow, from deduplication to extraction and risk of bias. SyRev focuses on AI title/abstract screening, with evidence you can report, and now adds data extraction and risk of bias and quality appraisal. Many teams use both: SyRev to remove clear exclusions against their criteria, then Covidence for dual screening, extraction and risk of bias.

Using SyRev before Covidence

  1. Search. Run your database searches as usual and export the results as RIS, PubMed .nbib, EndNote XML or CSV.
  2. SyRev pre-screen. Upload them to SyRev with your eligibility criteria: duplicates are merged, and at Careful SyRev auto-excludes only records whose estimated chance of inclusion is below 0.1%, logging the per-criterion scores behind every exclusion.
  3. Covidence. Export an RIS file with SyRev’s decision tagged on every record, and import the records SyRev did not exclude into Covidence for your usual dual screening.

Your methods paragraph, decision log and validation pack (a blind sample of the exclusions with a Wilson confidence interval) stay in SyRev for your write-up.

SyRev exports RIS files you can import into Covidence. SyRev is not affiliated with or endorsed by Covidence. Covidence is a trademark of its owner.

Try 2 free title/abstract evaluations

How the two tools differ (as of 25 September 2026)

As of 25 September 2026
SyRevCovidence
Price2 free title/abstract evaluations, then Replay $149 or Plus $249 per review · Team $29/seat/mo (fair use) · Department $2,500/yr (up to 10 named users) · University $9,000/yr (up to 150 named users) · Large from $15,000/yr · academic prices (commercial pricing)Free trial with 500 records · single-review and multi-review plans, see Covidence’s pricing page · organisation-wide licences on enquiry · concessions for teams in low- and middle-income countries[1]
AI screeningScores each title/abstract against your criteria; Careful mode auto-excludes only clear exclusions, and people screen the restActive-learning "most relevant" sorting, where people still screen all references · RCT classifier can tag records or remove likely non-RCTs, for medical and healthcare reviews[3][4]
Validation evidence96.5% recall at Careful on public benchmark reviews; individual reviews can be lower, often because of criteria wording · per-review validation pack: blind random sample of auto-excluded records with a Wilson confidence interval, decision-log CSV, methods paragraph and AI-use statementRCT classifier with high reported sensitivity (recall), evaluated on English-language biomedical records; per-feature validation and reporting templates[3][4]
PRISMAPRISMA 2020 diagram with AI and human exclusions shown separately, plus reports not retrieved and full-text exclusions with reasons and counts · SVG/PNG/PDF export · New PRISMA-S search log (every plan): each imported file logged, database and platform detected when the file names them, date, search string and record count per search; fills the Databases box and methods text; PRISMA-S CSV exportPRISMA flow diagram built from your screening decisions; updates automatically to show automated RCT-classifier exclusions[2][4]
Full-text screeningNew Human full-text screening on every plan, including the free evaluations: vote with the PDF open, single or dual, with a structured reason for every exclusion from a list your team can edit; conflicts resolved as at title/abstract · AI full-text screening on Plus, as second reviewer or suggestion · Plus: automatic open-access PDF fetch via Unpaywall and Europe PMC when records have a DOI, PMID or PMCID; you upload the rest, matched by DOI, PMID, filename or fuzzy title · every plan: “Find at my library” links through your library’s OpenURL link resolver or LibKeyFull-text screening; automatic upload of open-access PDFs via Unpaywall using the DOI[5]
Dual reviewAI as second reviewer (every plan, including free evaluations), or two human reviewers (Team and institutional plans) · blinded until both have screened: AI decision, other reviewer’s votes, exports and PRISMA diagram hidden · two human reviewers can be named, any two workspace members, or a reproducible random split across a reviewer pool · resolve step for disagreements · raw agreement and Cohen’s kappa reportedIndependent dual screening, blinding and structured conflict resolution as standard[2]
DeduplicationImport RIS, PubMed .nbib, EndNote XML or CSV, up to 10 files at once, or search PubMed, OpenAlex and Europe PMC from inside SyRev (every plan) · deduplication automatic across all files, before screening (every plan): clear matches (exact DOI/PMID/PMCID, or near-identical title with the same year and first author) merged; possible matches go to a check screen (merge or keep both); merges can be undone; removed records are not screened or counted toward usage; the count feeds the PRISMA 2020 diagramDuplicates removed automatically on import[2]
Data extractionNew Manual extraction on every plan, including the free evaluations · with Plus: AI pre-fill and an AI second extractor with a compare-and-resolve step, up to 500 studies per review, with an exact quote and page number on every suggested answer · two human extractors on Team and institutional plans · export CSV, Excel, a blank form or a RevMan Web pack, with change history and methods textReusable extraction templates, with LLM suggestions that must be verified[2][3]
Risk of biasYes New · RoB 1 (Higgins et al., BMJ 2011), AMSTAR 2, SYRCLE’s risk of bias tool for animal studies and the six NHLBI study quality assessment tools* · free on every planRisk of bias and quality assessment alongside extraction[2]

* SyRev is not affiliated with or endorsed by the National Heart, Lung, and Blood Institute (NHLBI) or the National Institutes of Health (NIH).

Where Covidence is strong

  • Independent dual screening, blinding and structured conflict resolution as standard[2].
  • The full workflow in one place: deduplication, screening, extraction, risk of bias and PRISMA[2].
  • A transparent, cautious approach to AI, with published validation figures and reporting templates[3][4].
  • Automatic open-access PDF upload at full-text stage[5].
  • Not-for-profit, with concessions for teams in low- and middle-income countries[1].

A note on AI auto-exclusion

In a public blog post, Covidence explains why it chose not to release criteria-based LLM auto-exclusion: performance "depended heavily on how eligibility criteria were written"[4]. SyRev's answer is a conservative default (Careful) and a blind validation sample on every review.

We think that's a fair description of the risk. SyRev's own testing points the same way: 96.5% recall at Careful on public benchmark reviews, with some individual reviews lower, often because of criteria wording. SyRev's answer is not a guarantee but a check on every review: a conservative default (Careful), and a validation pack with a blind random sample of auto-excluded records and a Wilson confidence interval, so you can see how your review behaved.

When adding SyRev helps

Adding SyRev helps when you want clear exclusions removed against your own criteria before records reach Covidence, with a per-review blind validation sample (with a Wilson confidence interval) you can report. On public benchmark reviews SyRev reaches 96.5% recall at Careful; individual reviews can be lower.

Dual review (AI as second reviewer on every plan; two human reviewers on Team and institutional plans) and import in RIS, PubMed .nbib, EndNote XML or CSV with automatic deduplication are available now; see features. Data extraction is newly launched (see data extraction), and so is risk of bias and quality appraisal (see risk of bias).

Try 2 free title/abstract evaluations See all features Back to comparison
Sources, checked 25 Sep 2026
  1. Covidence pricing: https://www.covidence.org/pricing/
  2. Covidence product overview: https://www.covidence.org/
  3. Covidence: overview of automation (AI) features: https://support.covidence.org/help/overview-of-all-automation-ai-features-available-in-covidence
  4. Covidence blog: How Covidence decides what AI to release (and what to hold back): https://www.covidence.org/blog/ai-screening-automation-systematic-reviews/
  5. Covidence: automatic upload of open access full texts: https://support.covidence.org/help/automatic-upload-of-open-access-full-texts

Competitor details come from their public websites on the date above and can change. If something here is out of date, tell us and we'll correct it.