AI can shave days off a literature review — and it can also invent five entirely fictional sources that look exactly like real journal articles. The skill is using it for the parts where it is excellent (organising, comparing, synthesising) and never for the part where it is dangerous (inventing citations). This tutorial draws that line clearly.
Anyone who has written a dissertation knows the slow torture of literature review week — twenty open tabs, four citation managers, and a growing suspicion that you are missing the one paper everyone else cites. AI can collapse a lot of that into a few hours of structured work. But it has a notorious weakness: when asked for sources, it sometimes invents plausible-looking citations that do not exist. Real-sounding authors, real-sounding journals, plausible page numbers — but the paper is fictional. Students have been failed, lawyers have been sanctioned, and academic careers have been damaged by this exact failure mode.
The fix is not to avoid AI for research. It is to use it for the tasks where hallucination is harmless (synthesising sources you provide) and never for the tasks where hallucination is catastrophic (inventing the sources themselves).
Think of a literature review as a funnel. At the top you have a broad topic. You narrow it down through search terms into a longlist of papers. You read those and reduce them to a shortlist. You then synthesise the shortlist into themes, identify the gap your work fills, and write the chapter. AI is genuinely useful at three of these five stages — and genuinely dangerous at one.
The clearest rule of thumb: you find the sources, AI helps you work with them. Use a real academic database — Google Scholar, Semantic Scholar, PubMed, JSTOR, your university library — to locate the papers themselves. Then use AI to help you read, compare, synthesise, and format.
Weak prompt — dangerous
Give me 10 academic sources on the
effects of microplastics on marine life,
in APA format.
AI will produce a beautifully formatted list of ten references. Some will be real. Some will be entirely fabricated — author names that exist, attached to papers they never wrote, published in real journals on dates when those journals did not have an issue. If you cite even one of these without checking, you have committed academic misconduct without realising.
Strong workflow — three safe prompts
--- PROMPT 1: search-term brainstorm ---
Act as a research librarian. I am researching
"the effect of microplastics on marine
zooplankton populations" for a Master's
dissertation.
Give me 8 search-term combinations I should
use in Google Scholar. Include synonyms,
adjacent fields, and Boolean operator syntax.
Do NOT give me any specific paper titles or
authors.
--- PROMPT 2: synthesise papers I provide ---
I have read the following 5 papers (full
references and abstracts pasted below). Act
as a literature-review assistant.
1. Cluster them into themes.
2. For each theme, list which papers support
which finding.
3. Note where papers disagree.
4. Identify the gap in the literature — what
has nobody studied yet that would naturally
come next.
Papers:
"""
… paste 5 full references + abstracts here …
"""
--- PROMPT 3: format a citation I already have ---
Format this reference in APA 7th edition:
Author(s): Sarah Chen and Maria Lopez
Title: Microplastic ingestion in copepods
Journal: Marine Pollution Bulletin
Year: 2023, Volume 184, pages 114210
DOI: 10.1016/j.marpolbul.2023.114210
Notice the discipline: AI suggests search terms, AI synthesises papers you have already verified, and AI formats citations from data you supply. At no point does AI invent a source.
Tip: The simplest test for whether you are using AI safely in research is this — could a strict supervisor or honour-code officer be in the room watching what you do, and you would have nothing to hide? If yes, you are doing it right.
Pick a topic you are researching (a paper, an essay, a dissertation chapter). Use Prompt 1 to generate eight search-term combinations. Run two of them in Google Scholar. Note how many papers you find that you would not have found with your own initial keywords.
Take five papers you have already read for a current project. Paste their references and abstracts into AI with Prompt 2. Compare AI's theme clustering and gap analysis to your own — what did AI notice that you missed? What did AI claim that does not match your reading of the papers?
Run a safety test: ask AI for five sources on a niche topic in your field, then try to locate each one on Google Scholar or your university database. Count how many actually exist with the exact title and author. The result will teach you, viscerally, why you should never trust AI for source discovery.
Sign in to join the discussion and post comments.
Sign inAdvanced Prompt Engineering Techniques
Master the powerful techniques AI experts use every day. Chain-of-thought, RAG, agents, function calling, prompt evaluation, and much more — 20 deep-dive tutorials.
Foundations of Prompt Engineering
The must-know basics of prompt engineering. Learn what prompts are, how AI models read them, and how to write clear instructions that get great results.
Prompt Engineering for Image Generation
Turn words into stunning visuals. Master AI image generation tools like Midjourney, DALL·E 3, and Stable Diffusion with 18 focused tutorials — from first prompt to full brand identity.
Prompt Engineering for Business & Productivity
Use AI to work smarter — automate tasks, make better decisions, and communicate professionally. 12 practical business prompt tutorials for professionals.
Prompt Engineering for Specific AI Tools
Tool-by-tool mastery — deep dives into ChatGPT, Claude, Gemini, GitHub Copilot, Midjourney, Stable Diffusion, and more. Learn the exact prompting techniques each platform rewards.
Prompt Engineering for Developers
Use AI as your coding co-pilot. 18 tutorials on writing prompts to generate clean code, debug faster, write tests, build APIs, and ship better software.