Skills on AI

Active theme: Light

Literature Review

Use when asked to write or conduct a literature review — a systematic survey and synthesis of existing published work on a topic, done before or alongside new research — as distinct from a [[research-proposal]], which proposes new work and often uses the review's findings to justify why that new research is needed.

A literature review is a systematic survey of existing published work on a topic, organized to show what's already known, where sources agree or disagree, and where gaps remain. It's not a summary of individual papers one after another — it's a synthesis that puts sources in conversation with each other so a reader can see the shape of a field, not just a list of what's been written about it.

Key components

  • Search strategy and scope — the databases or sources searched, the search terms used, the date range covered, and the inclusion and exclusion criteria for what counts as relevant. Recording this makes the review reproducible rather than an ad hoc pile of whatever the author happened to read.
  • Synthesis organized by theme or finding — sources grouped and discussed by the ideas, methods, or findings they share or dispute, not listed source-by-source in the order they were found.
  • Identification of gaps and contradictions — an explicit statement of what the existing literature hasn't settled, where studies conflict, and what hasn't been studied at all. This is often the actual point of doing the review: it's what justifies doing more work.
  • Critical evaluation — some assessment of the strength of the evidence behind each theme (sample size, method, replication) rather than treating every source as equally authoritative.
  • A clear scope statement up front — what topic, population, timeframe, or type of study the review covers, so a reader can judge what it does and doesn't claim to survey.

Why synthesis by theme beats a list of summaries

A reader of a literature review needs to understand how the sources relate to each other, not just what each one individually said. A sequence of "Smith (2019) found X. Jones (2021) found Y. Lee (2022) found Z." tells the reader nothing about whether X, Y, and Z agree, build on each other, or contradict outright. Organizing by theme — grouping sources around a shared question, method, or finding and discussing them together — is what actually shows the reader where consensus exists, where the field disagrees, and where the open questions sit. That structure is also what makes the gap analysis possible: a gap is visible only once related sources are placed next to each other.

Common pitfalls

  • Sequential summaries with no synthesis — a "review" that's really a list of individual paper summaries in the order they were read, with no sentence connecting how any of them relate to each other.
  • Search scope too narrow or too informal — relying on whatever turns up first in a casual search, rather than a defined strategy, risks missing directly relevant work — including work that contradicts the sources that were found.
  • Gaps stated too vaguely — "more research is needed" doesn't justify anything specific; a useful gap statement names exactly what question, population, or method hasn't been addressed, specific enough to point directly at a next research step.
  • Treating all sources as equally credible — citing a single small unreplicated study with the same weight as a large well-replicated one, without noting the difference in strength of evidence.
  • No stated inclusion/exclusion criteria — leaves a reader unable to judge whether the review missed something on purpose or by accident.
  • Outdated without acknowledgment — presenting a review as current when the search was done years earlier, without noting the cutoff.

Learn more

  • Research Proposal for proposing new research, often justified by the gaps a literature review identifies.
  • Survey Design for designing a new data-collection instrument once the review shows what's not yet known.
  • User Interview Guide for gathering qualitative evidence that published literature doesn't yet cover.

View literature-review/SKILL.md on GitHub