Learning & ResearchAdvanced1-3 days per review

Research Synthesis

The biggest risk in AI-assisted reviews is fabricated citations. This skill requires every conclusion to trace to a real paper and uses an evidence table to manage conflicting findings.

Workflow

  1. 1

    Define question and inclusion criteria

    Define the research question, date range, population, intervention and exclusion criteria. Record the protocol before seeing results to avoid moving the goalposts.

  2. 2

    Search and screen real papers

    Use academic databases and AI search tools for candidates, then open each paper to verify title, authors, year and DOI. Every AI-generated citation must be checked against the source.

  3. 3

    Build an evidence table

    Record sample, method, main result, limitations and relevance for each paper. Put supporting, opposing and null results in the same table.

  4. 4

    Write a review with tension

    Organize by theme, not paper-by-paper. Identify where evidence agrees, conflicts and is missing, then state your judgment and suggested future research.

Recommended prompts

FAQ

Can AI write a literature review?

It accelerates search, summarization and structure, but cannot make inclusion judgments or weigh evidence for you. Every citation must be checked against the source.

How many papers are enough?

It depends on the field and question. Systematic reviews often use dozens to hundreds; narrative reviews can use fewer but must cover major conflicts and gaps.

Related skills