Screen records for a systematic review

Any laptop with Python 3.10 or newer and 8 GB RAM. No GPU required for the standard ASReview workflow.

Component	Role	Price	Link
ASReview LAB 3.0.8	active-learning-assisted title and abstract screening	Free / 免费Apache-2.0 open-source softwarechecked 2026-08-02	https://github.com/asreview/asreview/releases/tag/v3.0.8
Zotero 9.x current stable channel	deduplication, full-text collection and citation export	Free / 免费desktop reference manager is free; 300 MB Zotero file storage is includedchecked 2026-08-02	https://www.zotero.org/download/

Launch a pinned ASReview project

uv venv --python 3.12 .venv
. .venv/bin/activate
uv pip install 'asreview==3.0.8'
asreview lab
# In the browser, create a project and import the deduplicated RIS or CSV export from Zotero. Export decisions after every screening session.

Screening decision schema

review_id: replace-with-protocol-id
record_key: DOI_or_stable_import_id
decision: include|exclude|maybe
exclusion_reason: population|intervention|comparator|outcome|study_design|publication_type|language|date|duplicate|other
reviewer: initials
decided_at: ISO-8601
notes: free_text
rule: Never use 'other' when a protocol-defined reason applies.

Known pitfalls

• Active learning changes screening order, not inclusion criteria.
• Seed the model with both relevant and irrelevant labels before trusting prioritization.
• Do not stop solely because the next records look irrelevant; use a preregistered stopping rule.
• Export decisions and project state regularly.

Browser only. Uploaded citations, abstracts and documents are processed by a cloud service.

Component	Role	Price	Link
Elicit Pro Pro plan current 2026-08-02	dedicated systematic-review workflow and large screening sets	$39/moofficial page states $480 billed annually while displaying $39 per user/month; arithmetic differs by $12 annually, so confirm checkoutchecked 2026-08-02	https://elicit.com/pricing
Zotero 9.x current stable channel	canonical library and final citation export	Free / 免费desktop software and 300 MB storage tierchecked 2026-08-02	https://www.zotero.org/

Protocol-bound screening instruction

Apply only the supplied eligibility protocol to titles and abstracts. For each record return include, exclude, or maybe; cite the exact protocol clause; quote the minimum text supporting the decision; and name one missing fact when the decision is maybe. Never invent a population, intervention, outcome or study design. Do not exclude on an unstated criterion.

Known pitfalls

• Elicit output must be independently checked against the review protocol.
• The pricing page has a small displayed monthly-versus-annual arithmetic mismatch; verify checkout.
• Keep stable record identifiers when moving between Elicit and Zotero.
• AI screening does not remove the need for duplicate human review when the protocol requires it.