Skip to content

Workflow Guide

AI4Meta follows the standard systematic review pipeline: Protocol → Codebook → Screening → Extraction → Analysis → Report.

1. Protocol

Define the research question, eligibility criteria, search sources, and analysis plan before operational work begins.

  • Question framework — record the appropriate PICO, PCC, corpus, or coding-frame structure for your project type.
  • Eligibility criteria — make inclusion and exclusion rules explicit before screening.
  • Analysis intent — note whether the review expects narrative synthesis, meta-analysis, coding, or mapping.

2. Codebook

Turn the protocol into auditable fields that reviewers can apply consistently.

  • Variables — define data type, allowed values, and reviewer instructions for each extraction field.
  • Coding rules — keep categorical labels and edge cases in one shared rubric.
  • Pilot pass — test the form on a small set of included studies before scaling up.

3. Screening

The screening phase filters imported studies based on your inclusion/exclusion criteria.

Title & Abstract Screening

  • Review each study's title and abstract.
  • Mark as Include, Exclude, or Maybe.
  • Use keyboard shortcuts (I, E, M) for rapid decisions.
  • AI suggestions appear as colored indicators — accept or override them.

Full-Text Screening

  • Upload PDFs for studies that passed title/abstract screening.
  • Review full text side-by-side with the screening form.
  • Record exclusion reasons for audit trail.

Conflict Resolution

When two reviewers disagree, conflicts are flagged for a third reviewer or consensus discussion.

4. Data Extraction

Extract structured data from included studies using the extraction matrix.

  • Codebook — define your extraction fields (study design, participants, interventions, outcomes, etc.).
  • Matrix view — spreadsheet-style data entry with one row per study.
  • AI auto-extraction — let AI pre-fill fields from PDFs; review and confirm.
  • Validation — built-in checks for missing data, out-of-range values, and consistency.

5. Analysis

Run statistical analyses directly in AI4Meta:

  • Select outcome data and effect measure (OR, RR, MD, SMD, etc.).
  • Generate forest plots, funnel plots, and more.
  • Run subgroup, sensitivity, and meta-regression analyses.
  • See the Analysis Guide for details.

6. Report Generation

Generate publication-ready reports:

  • PRISMA flow diagram — auto-generated from screening data.
  • Summary of findings table — with GRADE assessments.
  • AI-written narrative — draft methods and results sections.
  • Podcast generation — turn the saved report into a solo briefing, two-speaker podcast, or three-speaker panel with downloadable audio.
  • Export — Word (.docx), PDF, or LaTeX formats.

PRISMA Flow Diagram

AI4Meta automatically tracks the number of records at each stage and generates a PRISMA 2020-compliant flow diagram. Customize labels and export as SVG or PNG.