AI in literature review workflows

Understand how generative AI may be used for literature reviews from topic scoping to literature discovery and screening and what you must consider to ensure rigour and integrity in your work.

Responsible GenAI use

Whilst generative AI (GenAI) could improve ease and efficiency of literature reviews, you must use it in a way that does not compromise the rigour of your review methodology or research integrity.

Considerations for GenAI across different review methodologies

Different literature review methodologies require varying levels of rigour. However, all review types aim to be unbiased, transparent and replicable. Introducing genAI tools into review workflows can pose a risk to these principles. Outputs are stochastic, random and probabilistic, not consistent every time, limiting replicability. Tool algorithms are hidden or proprietary, limiting transparency around their workings. Outputs can be biased by various factors, including information accessibility. This is particularly relevant in literature reviews, where access to non-digitised or paywalled data can be a limitation.

As such, for highly rigorous methodologies like systematic reviews or meta-analyses, genAI use may be limited to specific steps or assistive genAI integrated into traditional workflows, rather than full automation.

GenAI use in review workflows

GenAI can support different stages of a literature review, but outputs must always be verified. Some potential applications could be:

  • Initial scoping and topic development – Use AI to explore recent literature, key publications and research gaps, get to know a new area and start collecting seed articles.
  • Database search strategy development – Get an initial search strategy to refine, then troubleshoot it to get optimal results, then translate your strategy across your target databases.
  • Literature discovery – Use natural language queries to search for articles or use seed articles to build a citation and semantic search network for hand searching.
  • Screening results – Use validated automated screening tools with subsequent human verification or assistive tools that make manual screening more efficient or easier.
  • Abstract data extraction – Use language models to pull key information from article abstracts, then verify against the original materials.
  • Synthesis – Use closed University-approved tools to assist in structuring your ideas and to suggest additional points to consider when writing your synthesis.

Copyright warning: Tools that require uploading full-text articles for analysis will violate the terms of the University's subscription database licenses and may risk our access to these databases. Refrain from full-text uploads of copyrighted materials.

GenAI bias and hallucinations in literature reviews

All steps above are vulnerable to genAI biases and hallucinations:

  • Access biases: GenAI's limited access to paywalled or non-digitised research will limit the information presented in outputs, despite the appearance of being authoritative. You must seek additional sources of primary materials, such as through library databases, alongside AI outputs, to be comprehensive.
  • Cultural or stereotyping biases: Models tend to reflect Western or Eurocentric viewpoints and may perpetuate harmful pre-existing stereotypes. Critically evaluate cultural commentary and consult a broader range of international and indigenous sources.
  • Methodological biases: GenAI can favour dominant methodological approaches and underrepresent less common or emerging methods. Additional literature searching alongside GenAI tools is needed to ensure methodological diversity.
  • Hallucinations: GenAI can fabricate citations and misrepresent findings. Validate all references and claims against original sources before including them in your review.

Workshops and courses

AI Essentials

Learn to engage with AI effectively, efficiently, safely and responsibly.

Go to AI Essentials

AI for literature reviews workshop

Online workshop discussing benefits and risks of using AI in literature reviews for researchers, particularly postgraduates and doctoral candidates.

Go to AI for literature reviews workshop

Contact