Literature Review Software by Stage: A Map of What Each Kind Does

Your supervisor says Covidence, a colleague swears by Rayyan, the library teaches Zotero, and the person at the next desk does everything in Elicit. These are not competing answers to the same question, and nobody has said so.

Here is the short answer. A review is a sequence of stages and most software is built for one of them: a database finds records, a reference manager keeps them, a screening platform decides which ones stay, and a writing tool drafts what you say about them. The names traded in corridors belong to different stages, and the trouble starts when one kind of tool is asked to do another kind's job.

The map: tools by stage

Each row is one tool and each column is one stage of a review in the order you meet them: search, manage references, screen, extract, appraise, analyse (mapping counts as analysis here), write, and report, which means the PRISMA flow diagram.

Legend: ✓ the vendor documents the stage · ◐ covers part of it · — not offered · n/d the vendor's pages do not say.

1.Literature search software

Search is where the vocabulary is loosest: a database, an open index, an AI assistant and a citation map all answer with a list of papers, and they are four different things.

The assistants and the citation tools are good at finding what you did not know to look for, and the database query is the one you report in the methods; most reviews need both.

2.Literature management tools

A reference manager holds the library: every record you export from a database, every PDF, and the citations you insert while you write.

The screening platforms import your export and most remove duplicates on the way in, and none of them replaces the reference manager: the library is still there when the review is over.

3.Systematic review screening software

Screening is the stage the review platforms were built for: the same records in front of two people, a decision from each, and a place to settle the disagreements.

Four of the platforms are compared feature by feature in the comparison post.

4.Data extraction software for systematic reviews

Extraction turns every included paper into the same set of answers: who was studied, the setting, the method, the findings, whatever your form asks.

In the platforms, appraisal is a special case of extraction and lives next door; in the assistants, the extraction table is what later becomes the report.

5.Critical appraisal

Two questions get tangled here, so keep them apart: which checklist to use follows the design of each included study, the subject of the appraisal post; this section is only about where you fill it in.

Pick the framework by study design first, then look at the second row: if your checklist is not on a platform's list, it goes in as custom questions, and the methods section is none the worse for it.

6.Literature analysis software and literature mapping software

Analysis means two jobs that share a word: coding what the included papers say, and mapping how the papers of a field connect to each other.

A statistical meta-analysis is its own job: if your platform does not run one, it happens in statistics software after extraction.

7.Literature review writing software

Writing is the stage the AI research assistants are really selling: give them the papers, get back a draft with citations. A few review platforms now draft from the review's own record instead.

A citation plugin is not a writer: Zotero, Mendeley and EndNote insert and format references in your manuscript, which is why they sit under managing references and not here.

8.Creating a PRISMA flow diagram

A systematic review reports its counts in the PRISMA 2020 flow diagram and a scoping review in the PRISMA-ScR version, and the only question is whether the numbers come from the software or from you.

If your review lives in a spreadsheet and a reference manager, the generator is the missing piece.

Which combination do you need

Three kinds of review, three sets of stages. The second column names kinds of tool rather than products, because the map already tells you which products cover which kind.

FAQ

Three questions come up every time.

Is Zotero a literature review tool? Zotero is a reference manager: it holds the records and the PDFs and puts the citations in your manuscript. It does not search databases, screen or extract. For a narrative review, Zotero and a document are often enough; the moment two people have to screen the same records, you need a screening platform beside it.

Can one tool do the whole review? Not from database to manuscript. Nested Knowledge comes closest, with a mark in every column but reference management; Elicit covers the AI-first path from search to report. Everyone else combines: a database, a reference manager, one platform for the middle stages, and a writing tool if wanted.

What is the difference between a reference manager and a screening tool? A reference manager keeps your library and your citations for as long as you do research. A screening tool holds one review's records for the length of that review, records each reviewer's decision, and produces the counts for the PRISMA diagram. You export from the first into the second, and the first is still there when the review is over.

If you are choosing tools this week: write down which of the eight stages your review has, pick one tool per stage from the map, and check that each hands its output to the next, an export on the way out and an import on the way in. Then write the names into your protocol. The rest is settings.