Frequently asked questions

You can frame the question, prepare the search and write criteria. Those features include a few trial calls without verification. But screening, full text, extraction, risk of bias, manuscript and assistant will not start until you verify the address. Check spam. If it is not there, request another link from the app without signing in. The link lasts 24 hours and works once. Waiting does not erase your work.

Use the quick start. Follow the path from question to export and open detailed pages as needed.

No. Once you approve Question, the question and framework are locked so the method cannot change halfway through. Create a new project for a different question.

No. Seven frameworks are available on every plan: PICO (intervention), PECO (exposure), PCC (scoping review), CoCoPop (prevalence), PIRD (diagnostic tests), PEO (prognosis) and SPIDER (qualitative research). The app suggests one based on your question and you can change it. See Question.

The whole path through to the manuscript. AI screens your first 100 records; the rest are marked for you to verify. AI full-text screening, extraction, risk-of-bias draft and manuscript draft are paid. On the free plan, you do those four manually with the same decision record. Export is always free on every plan. See plans.

Yes. Each project selects Spanish or English at the Question stage. That choice sets the language of criteria, verdicts and manuscript. You can also change the interface language.

Check the language filter and year range first. PubMed queries are usually in English, so try “All languages”. For broader coverage, run several variants and manually import results from SciELO, LILACS or Dialnet.

At Import studies, upload RIS, BibTeX, CSV, PubMed (.nbib) or Web of Science .txt, or paste exported content. Common databases and reference managers export RIS or BibTeX. References are deduplicated. Imports from recognised databases count under “identification via databases and registers”; items labelled “Other methods” use the corresponding PRISMA branch.

Unpaywall can retrieve the document only when a direct file link exists. Many open-access articles link only to their webpage. Use “Upload PDF” to add it manually.

Raise the lower threshold in “Full screening” to exclude more uncertain studies. At 0.5, uncertainty never causes exclusion. Raising it shortens the queue at the cost of accepting some additional exclusions. See how to adjust the band.

The model is not perfectly deterministic. By design, any variation ends in human review, never silent exclusion. Code calculates the decision from criterion verdicts.

No. It evaluates each criterion and code applies deterministic rules. Uncertain cases go to your queue, your decisions are audited and prevail, and no stage advances without approval.

Limits cover active reviews, AI budget, 50 searches every 24 hours and imports of up to 5 MB per file. The number of AI-screened and stored records depends on your plan. The app shows the effective limit in “Project status”. On the free plan, OpenAlex allows three searches per review.

You can continue screening and making manual decisions, and nothing is lost. The budget limits actions that call the model, not your work or exports.

Yes, and export is always free. You can download references, screening, extraction and risk-of-bias tables, PRISMA 2020, a PROSPERO protocol, the audit log and a manuscript draft. See PRISMA and export.