Hour 1 · SLR · Step 5 of 1533%

Screening — AI assists with relevance

The core of the SLR.

~15 min

Two tools dominate AI-assisted screening. Both learn from your include/exclude decisions and re-rank the rest of the pile so you read the most-likely-relevant records first.

Tools for this step

Rayyan

Web-based · friendly · best for workshops & small teams

•Free tier covers 3 active reviews
•Learns from your include/exclude calls
•Rates every record with a relevance score

ASReview

Open-source · free · best for large reviews

•Active learning prioritises records
•Cuts screening effort by ~70–90%
•Fully transparent and reproducible

AI prompts (1)

Prompt

Title & abstract screening assistant

When: You're unsure about a borderline record and want a structured second opinion.

You are helping me screen a record for a systematic literature review.

Research question: <PASTE>
Inclusion criteria: <PASTE>
Exclusion criteria: <PASTE>

Record:
Title: <PASTE>
Abstract: <PASTE>

Do not give a yes/no in isolation. Instead:
1. For each inclusion criterion, mark MET / NOT MET / UNCLEAR with a one-line justification grounded in the abstract.
2. For each exclusion criterion, mark TRIGGERED / NOT TRIGGERED with a one-line justification.
3. Give a final recommendation: INCLUDE for full-text / EXCLUDE / NEEDS FULL-TEXT TO DECIDE.
4. Note any claim you had to infer rather than read directly in the abstract.

Never let the model be the final screener. It's a second pair of eyes; the decision is yours, and conflicts are resolved by a human reviewer.