How to write a literature review with AI, step by step
To write a literature review with AI, let it speed up the slow parts, finding, sorting, and summarising real papers, while you keep the synthesis, the judgement, and the writing. Used that way it saves real time; used as a ghostwriter it invents citations and tanks your credibility. Never let it generate your reference list, and verify every source against a database before it goes in.
A literature review is one of the highest-stakes things you will write in school, because it is almost entirely citations. That makes it the worst place to lean on a chatbot's memory and the best place to use AI carefully. This is a step-by-step way to do that: narrow the topic, find real sources, screen them, verify them, sort them into themes, then synthesize and write. AI helps at every step. It just never gets to decide what is true or what exists.
What a literature review actually has to do
Before you automate anything, be clear on what you are grading yourself against. A literature review discusses published information in a subject area and is organized around ideas and themes, not source by source the way an annotated bibliography would be. The point is not to list what each paper said. The point is to show how a field of work fits together and where your own project sits inside it.
That distinction is the whole game. The skill graders are looking for is synthesis: a strong review draws on several sources, shows how they relate to one another, and makes the writer's own point, instead of describing each study in turn. A chatbot can produce fluent paragraphs, but the argument that connects the studies has to be yours. Outsource that and you have outsourced the assignment.
Use AI as an assistant, not an author
University libraries are direct about where the line sits. Texas A&M-Corpus Christi tells students to approach these tools as research assistants to work more efficiently, and never to rely on AI to do the work for you, partly because the same guide notes that AI tools are notorious for hallucinating citations that do not exist. Duke's library makes the same point: AI does not replace searching library databases like Web of Science or PubMed, and you should always verify and read your sources.
So divide the labor honestly. Good jobs for AI:
- Brainstorming search terms and synonyms you would not have thought of
- Summarizing a paper you have actually downloaded, so you can triage faster
- Suggesting how to group studies into themes
- Tightening prose in a paragraph you already wrote and checked
Jobs you keep:
- Deciding which sources are credible and relevant
- Reading the sources you cite
- Building the argument that links them
- Writing and owning the final words
The six steps, with AI in the right places
Most university guides describe the same workflow. Niagara University lays it out as a repeatable sequence: select, search, evaluate, analyze, synthesize, present. Here is each step with the place AI helps and the place it cannot.
Narrow your scope
Pick a question you can actually cover. The narrower your topic, the easier it is to limit how many sources you must read to survey the field. AI is genuinely useful here: ask it to break a broad topic into narrower sub-questions, then choose one.
It is worth seeing what narrowing is actually worth, because it is not a rounding error. On 11 August 2026 we counted journal articles published since 2021 in the open OpenAlex index, searching titles and abstracts. "Microplastics" returns 34,406 of them. Add one word, "microplastics rivers", and it is 3,172. Add one more, "microplastics freshwater rivers", and it is 803. Two words cut the reading pile by 98 percent. The same shape holds elsewhere: "urban heat" gives 18,530, "urban heat island" 9,265, and "urban heat island tree canopy" 259.
And the pile is growing under you. Articles with "microplastics" in the title or abstract went from 22 in 2005 to 43 in 2010, 210 in 2015, 2,233 in 2020 and 8,838 in 2025. A field can publish more in one year than in its entire first decade, which is why a scope that was reasonable for someone's thesis five years ago may not be reasonable for yours. Every count here comes from the public OpenAlex works endpoint, so you can run it on your own topic before you commit to it.
How many sources you land on is a question for your supervisor, but a published figure helps you argue with yourself. The University of Southampton library, writing about dissertations, offers a "ballpark figure" rather than a rule: for a standard dissertation of around 10,000 words, "referencing 30 to 40 sources in your literature review tends to work well" (Writing the dissertation). Set against the counts above, that is the real job: 30 to 40 out of tens of thousands, chosen on purpose.
Search for real papers
Use AI to generate keywords, Boolean strings, and author names to look for. Then run the actual searches in a real database or a tool that retrieves real literature. This is the step where people get burned: if you ask a general chatbot for the papers themselves, it will often invent them. The search has to hit something real, not the model's training data. If your review needs to rest on refereed work, how to find peer-reviewed articles walks through where to search and how to confirm a paper was actually peer-reviewed.
Screen and evaluate
Skim each result against your scope and a simple quality bar (peer-reviewed, recent enough, relevant). AI can draft a one-line summary of a paper you have downloaded to help you triage, but you decide what stays. Discard aggressively. A focused review built on twenty strong sources beats a sprawling one padded with forty weak ones. For recency, a useful rule of thumb is to favor work from the last 5 to 10 years, and to reach further back only for foundational papers that newer studies keep building on. If you want a named method to lean on, look up the CRAAP test (currency, relevance, authority, accuracy, purpose) or the SIFT method for checking sources.
Verify any reference an assistant suggests exactly as you would your own finds. There is no category of source that gets a pass for having come from a tool you like.
Verify every citation
Before any reference enters your draft, confirm it is real. The checks take seconds each:
- Search the exact paper title in Google Scholar or your library catalog
- Paste the DOI after
https://doi.org/and confirm it resolves to that paper - Google the lead author to confirm they exist and publish in the field
Do not skip this because a link works. We cover the full method in how to check if a citation is real.
Sort into themes
Lay your verified sources out and group them by the ideas that connect them: studies that agree, studies that conflict, the gap nobody has filled. This is where you stop collecting and start reviewing. AI can suggest candidate groupings from your summaries, but read them critically; the themes are your map of the field.
Synthesize and write
Write each theme by putting sources in conversation, not in a line. "Three studies found X, but Smith's larger sample found the opposite, which suggests Y" is synthesis. "Smith found X. Jones found Y. Lee found Z" is a summary wearing a review's clothes. Draft with AI if you like, then verify every claim against the source it rests on and rewrite it in your voice. To see what a synthesized theme reads like on the page, our worked literature review example walks through one with a reusable structure template.
Why verification is a step and not a caveat
Chatbots like ChatGPT are statistical prediction tools that produce plausible-looking text rather than systems that retrieve and check real sources, so a citation is just another plausible string to generate. Measured fabrication rates for the popular models run from roughly one in five references upward, which for a document that is mostly references is not a risk you can absorb. We set the numbers out, with the studies behind them, on the tool-buying page linked at the end, and the mechanism is in why AI makes up citations.
What matters for your method is narrower: one fabricated source is enough for a grader to question an entire review, and "the AI gave it to me" is not a defense when your name is on the page. So verification is not a warning attached to the workflow. It is a step inside it, with its own place in the order, which is why it sits above between screening and sorting rather than in a box at the bottom of this page.
Put the source before the sentence
There is one habit underneath everything on this page, and it is worth naming because it is the difference between a review that holds and one that quietly does not. Write in the direction the evidence flows. Find what the literature says, then write the sentence that reports it. Never write the sentence you want to be true and then go looking for a citation to hang on the end of it.
That second order feels productive and it is how most bad reviews get written, with or without AI. Once a sentence exists, you are no longer searching, you are shopping: you will accept a paper that is nearly on point, cite it for something adjacent to what it actually found, and move on. It is also exactly the order a general chatbot writes in, which is why its citations fail in the same way yours do when you work backwards. Reversing it costs nothing and removes the whole class of problem, because a claim written from a paper you just read cannot cite a paper that does not exist.
Here is where we would put the effort, and it is not where most guides put it. Given the choice between a better tool and a narrower question, take the narrower question every time. Look at the counts above: the difference between "microplastics" and "microplastics freshwater rivers" is 33,000 papers, and no assistant, however good, closes a gap that size in your favour. Scope is the only lever in this process with that kind of leverage, and it is free, and you pull it before you have spent an hour on anything. A tool applied to a question that is too broad just helps you produce a padded review faster.
If what you actually want is to choose a tool rather than run the method, that is a different job and it has its own page: AI literature review generator covers what these products can and cannot do, what the fabrication studies measured, current prices we checked ourselves, and where we would send you instead of us.