revisia Log in Sign up

Faster systematic reviews with artificial intelligence

The AI searches, screens and extracts the data. You decide every step with the article's exact sentence in front of you.

Step 1 · Your question, inside its framework

Evaluate the effects of metformin versus placebo on longevity in older adults without diabetes

  • Population older adults without diabetes
  • Intervention metformin
  • Comparison placebo
  • Outcome longevity

Here the framework is PICO. revisia also works with PECO, PEO, PCC, SPIDER, CoCoPop and PIRD, and turns these concepts into the search for PubMed, Europe PMC, OpenAlex, Semantic Scholar and ClinicalTrials.gov.

Refine your question with AI

  1. 1 Question inside its framework
  2. 2 Search 5 databases at once
  3. 3 Import all your databases, no duplicates
  4. 4 Screening the doubtful ones, for you
  5. 5 Data each value with its quote
  6. 6 PRISMA 2020 built from your counts
  • Word for word revisia looks up every quote in the article's text. If it cannot find it exactly, the cell is flagged “verify”.
  • All your databases Import what you export from Scopus, Embase, Web of Science or CINAHL. revisia removes the duplicates and the AI screens the lot.
  • Pay once You pay per review, with no subscription to renew. Exporting your work is always free.
The whole journey

Duplicates get filtered. Doubt does not.

revisia gathers your searches and your RIS, BibTeX, CSV, .nbib or Web of Science files. Then it deduplicates and applies the criteria without hiding the uncertain cases.

PubMed Europe PMC OpenAlex Semantic Scholar ClinicalTrials.gov You import Scopus WoS Embase CINAHL + anything else DUPLICATES CRITERIA to your review included excluded

Searches and imports go through the same process.

How it works

How to do a systematic review with revisia, in 6 steps

Each step explains what it is, what revisia does and what you decide. If this is your first review, start here.

  1. Write your research question

    What is PICO?
    The four parts of a clinical question: Population, Intervention, Comparison and Outcome.

    Write the question however it comes out. revisia splits it into its parts and suggests a sharper version.

    You decide the final wording and the framework: PICO, PECO, SPIDER and others.

    revisia's Question stage: “Evaluate the effects of metformin versus placebo on longevity in older adults without diabetes”, each PICO concept in its colour, and the list of the 4 concepts that will be searched
  2. Import from every database you have access to

    What is deduplication?
    Removing repeated records when the same study turns up in several databases, so you screen it only once.

    Export from Scopus, Embase, Web of Science, CINAHL or any other database and upload the file: RIS, BibTeX, CSV, PubMed .nbib, Web of Science .txt or WHO ICTRP. revisia recognises the format on its own.

    What you import joins what you searched in step 2. revisia removes duplicates by comparing DOI, PMID and title, and the AI screens everything together.

    You decide whether two records are the same study: every merge is logged with its reason and you can split it again.

    Import studies stage: 102 articles in the project, 11 matched across databases, 2 searches and imports, and the drop zone for RIS, BibTeX, CSV, PubMed .nbib, Web of Science .txt or WHO ICTRP files
  3. Screen the articles: the AI proposes, you decide

    What is screening?
    Reading each study's title and abstract and deciding whether it meets your inclusion criteria.

    The AI answers criterion by criterion: yes, no or uncertain, with the sentence from the abstract that backs it. If the abstract is not enough to decide, the study stays pending, in your queue and not in the bin.

    A real case: for the ANTHEM trial the AI marks all five criteria “yes” and quotes the sentence for each. For the population: “older adults without diabetes”. Its score is 1.00: it meets every criterion (the score runs from 0 to 1 and an “uncertain” counts as half).

    You decide every doubtful case. If you change an AI decision, you note the reason and it is recorded.

    The ANTHEM trial included by the AI with 1.00: all five criteria “yes”, each with its sentence from the abstract
  4. Extract each study's data, with its quote

    What is data extraction?
    Putting what you need from each included study into a table: sample size, intervention, results.

    revisia fetches each study's open-access PDF; you upload the rest. Then it fills in your columns, and every value carries the sentence from the article it comes from.

    You decide the columns and check the cells flagged “verify”. If the article does not say, the cell reads “Not mentioned”.

    Evidence table for two metformin trials: sample size, age range, follow-up and mortality “Not mentioned”
    Age range and population criteria taken from the full text, with the article's sentence highlighted and the button to open the PDF on page 2
  5. Assess risk of bias and download the PRISMA 2020 diagram

    What is PRISMA 2020?
    The reporting guideline for systematic reviews. Its flow diagram shows how many studies come in and drop out at each stage.

    Choose the risk of bias tool: RoB 2, ROBINS-I, QUADAS-2, NOS and more. The AI drafts each domain, with its quote. The PRISMA diagram draws itself from your project's real counts.

    You decide every risk of bias judgement. The diagram downloads as SVG or PNG.

    PRISMA 2020 flow diagram generated by revisia from the project's counts
    RoB 2 traffic-light risk of bias figure

Type your question and try revisia for free

Create project

What you gain

Less mechanical work. More control over the evidence.

01

Start reading without duplicates

revisia filters duplicates on import. Every merge keeps its reason and its source.

02

Lose fewer studies

Anything doubtful lands in your review queue. It is never discarded on an assumption.

03

Cut screening from days to minutes

A thousand abstracts are screened in about two minutes. You review what is left standing.

04

Spot the data that is missing

If the text does not say it, the cell reads "Not mentioned". It is never confused with "not extracted".

05

Work in English or Spanish

Each project picks its language; criteria, verdicts and the manuscript come out in it. Every study keeps its original text.

06

Defend every data point

Every extracted value carries the sentence it came from. You can also export the decision log.

Every number on this page was measured and counted. Full methodology →

Who it is for

Faculty and researchers

For your reviews and your group's: a reproducible method, results published alongside their methodology and an audit trail that stands up to peer review.

PhD candidates

The whole process, guided through stages that cannot be skipped, with PRISMA 2020 and a manuscript draft to start your methods chapter.

Master's and undergraduate theses

Your first systematic review without getting lost: two-phase criteria, doubt always in your hands and deliverables ready to defend, with no paywall on your work.

Also for librarians and information specialists (versioned, auditable strategies) and clinical and health groups (two-phase screening, open-access PDFs, extraction with verified quotes).

What revisia does NOT do

The real limits, stated before you run into them.

It does not search Embase, Scopus, Web of Science, CINAHL or Cochrane

Native search covers PubMed, Europe PMC, OpenAlex, Semantic Scholar and ClinicalTrials.gov. The rest you search yourself and bring in as RIS, BibTeX or CSV: it joins the same pile and is deduplicated with everything else.

It does not translate the studies

You work in English or Spanish: criteria, verdicts and the manuscript come out in your project's language. You read each study's text in its own language. No machine translation in between.

It does not split a merge made by DOI

Every merge is on record: what was merged, why and from which source. The count feeds your PRISMA. If two records were merged because their titles matched and they are not the same study, you split them with a button. If they share a DOI, you cannot: they are the same article.

Free tools

Tools you can use today

Loose parts of a review you can solve right now, without an account: the figure or the document the journal asks for, ready in your browser. Your data never leaves it.

PRISMA 2020 flow diagram drawn by the tool, with the identification, screening and included boxes

PRISMA 2020 flow diagram

All four official templates, arithmetic checked as you type, and export to PDF, SVG, PNG and Word.

PRISMA 2020 checklist with the status and location of the first items filled in

PRISMA 2020 checklist

All 27 items plus the 12 abstract items with where each one is reported; saves as you go and exports to Word and PDF even half done.

Traffic-light risk of bias figure with seven studies and the five RoB 2 domains

Risk of bias figure

Traffic-light and summary plots for RoB 2, ROBINS-I, QUADAS-2 and Newcastle-Ottawa; paste the table from Excel.

All tools →

Frequently asked questions

What people usually ask

How much does it cost?

You start free, with no card. No subscription to renew or cancel: you pay for the review you do, with plans sized by the number of records, and that project's AI stays available for twelve months. Your work never expires: exporting it is free, always, and that is product policy, not a promise.

What is free and what is paid?

The free account takes you through the whole flow, up to the manuscript. The AI screens your first 100 records; the rest are marked for human verification. Full-text screening, data extraction, risk of bias assessment and the manuscript draft with AI are in the paid plans; on the free plan you do those stages by hand, and the PRISMA diagram, the audit trail and the exports are yours either way.

What happens to my data?

We only use an essential session cookie, no analytics or advertising cookies. Plausible measures visits and usage actions on the website and in the app. To compute daily metrics it processes the IP address and User-Agent, but it does not store those raw values; we never send it your identity, your texts or your reviews' data. You can ask for your account to be deleted. Policy and terms are in the footer (in Spanish).

Which languages does it work in?

Yes. Each project picks its language (English or Spanish), and that sets the language of the criteria, the verdicts and the manuscript. The interface is available in both.

Which sources does it cover?

Native search in PubMed, Europe PMC, OpenAlex, Semantic Scholar and ClinicalTrials.gov, all five on the free plan too, where OpenAlex allows three searches per review. For Web of Science, Scopus, Embase, CINAHL or Cochrane you import your RIS, BibTeX or CSV files and they are deduplicated with a record of every merge. We would rather explain the limits than hide them.

Does the AI decide for me?

No. The AI assesses each criterion and the code applies deterministic rules; uncertain cases go to your review queue, and your decisions are audited and take precedence.

Can I check where each data point comes from?

Yes. Every value in the evidence table carries its verbatim quote and its source (abstract or full text); a “maybe” that no sentence backs is marked “no quote”. If the study has a PDF, "Open PDF" jumps to the quote's page with the sentence highlighted. And when the code could not confirm a quote, the cell is flagged "quote not found" instead of being taken as good.

Can I export my work?

Yes: records, evidence tables, PRISMA counts, bibliography (RIS, BibTeX, CSV) and the manuscript draft. Exports are free, always.

Create your account and start today

From the research question to the PRISMA diagram, with every decision audited. Free to start.

Create a free account