Is it cheating to use AI for a literature review?

It depends on what you use it for and what your institution and assignment allow. Using AI-powered search to find papers, or to help you understand a difficult article, is consistent with every policy quoted in this article, and Stanford's default treats it like help from another person. Submitting AI-written paragraphs as your own analysis, without permission and acknowledgement, is academic misconduct under policies such as Cambridge's. Check the brief, and when it is silent, ask.

Do I have to declare that I used an AI tool to find sources?

Every policy quoted in this article asks for AI use to be acknowledged, and Stanford advises students to default to disclosing when in doubt. For search tools the declaration is short: name the tool and say what you used it for, in your methods section or in an acknowledgement. You cite the papers you found and read, not the tool. Elsevier exempts basic grammar, spelling and punctuation checks from its declaration for journal submissions; check whether your own institution does the same.

Can I cite ChatGPT or another chatbot as a source?

Not as evidence for a factual claim. The ICMJE states that referencing AI-generated material as the primary source is not acceptable, and chatbots can produce references to works that do not exist. If you quote or discuss AI output itself, for example in a study about AI, cite it in the format your style guide gives. Otherwise find, read and cite the original source.

Can my university detect AI use?

Detection software exists, and published tests question its accuracy. Weber-Wulff and colleagues (2023) tested 14 detection tools and concluded that they were "neither accurate nor reliable". Liang and colleagues (2023) found that GPT detectors frequently misclassified writing by non-native English speakers as AI-generated. The University of Sydney tells students that a detector score "would not be the only evidence relied upon". It is the wrong question to plan around. What helps if your work is questioned is a record of your process: saved searches, dated notes, drafts with version history and a log of any AI use. That record also makes your disclosure statement easy to write.

My supervisor says no AI at all. Does that include AI-powered search tools?

Ask, and be specific. "AI" covers everything from relevance ranking in a search engine to a chatbot that drafts text, so a blanket instruction can be read in more than one way. Your supervisor may mean no generative AI for writing or analysis, or may mean something broader. Name the tools you would like to use and what for, agree in writing which are acceptable, and keep the email.

Using AI in a Literature Review: What University and Publisher Policies Actually Allow

Students who want to use AI responsibly face a practical problem. University policies are written at the level of principle, course rules vary from one module to the next, and none of them mention "literature review" by name. This article reads the policies closely and maps them onto the actual tasks involved in reviewing literature.

Policies change. Everything quoted here was checked against the institutions' own pages in September 2026. Your own institution's current wording is what counts.

What the policies say

University of Cambridge

Cambridge permits AI for learning and draws the line at assessed work: "Students are permitted to make appropriate use of artificial intelligence tools to support their personal study, research and formative work." And: "A student using any unacknowledged content generated by artificial intelligence within a summative assessment as though it is their own work constitutes academic misconduct, unless explicitly stated otherwise in the assessment brief."

Two details matter. The offence is unacknowledged content presented as one's own. And the assessment brief can override the default in either direction.

Stanford University

Stanford's Board on Conduct Affairs set a default for courses whose instructors say nothing: "Absent a clear statement from a course instructor, use of or consultation with generative AI shall be treated analogously to assistance from another person." It adds: "Students should acknowledge the use of generative AI (other than incidental use) and default to disclosing such assistance when in doubt."

The "another person" test is the most useful single heuristic for students anywhere. Would it be acceptable for a knowledgeable friend to do this for you? A friend can recommend papers, explain a statistical method or point out that your argument has a hole in it. A friend cannot write your synthesis section.

University of Toronto, School of Graduate Studies

For theses, Toronto is stricter about process: "The use of generative AI tools in any aspect of researching or writing of the thesis must be done with the prior approval of the supervisor(s) and supervisory committee." The guidance warns that "Generative AI tools have also been shown to reference scholarly works that do not exist", reminds students that they "are ultimately responsible for the content", and raises a point students rarely consider: "Using AI to lessen the burdens of writing could undermine the development of these invaluable skills."

If you are a graduate student, the takeaway is to have the conversation with your supervisor before you start, not after the first draft.

University of Sydney

Sydney tells students: "you're allowed to use artificial intelligence (AI) tools as a general learning aid. However, there are limitations on how you can use these tools to complete assessments." It also instructs: "Ensure that you have acknowledged any use of software or tools, and that you reference all sources." In November 2024 the university announced that from Semester 1 2025 its default would allow AI in assessments other than exams and in-semester tests, unless teaching staff decide otherwise, and that it would then move to what it calls a "two-lane approach": secure, in-person assessments completed without external aids, and "open" assessments that support the use of all available and relevant tools, with students acknowledging how they use AI. Where AI use is permitted, the student guidance says to keep track of how you used generative AI and to keep copies of its outputs as evidence of your research and writing process.

The Russell Group (UK research universities)

The Russell Group's five principles, published on 3 July 2023, set the direction for UK policy:

Universities will support students and staff to become AI-literate.
Staff should be equipped to support students to use generative AI tools effectively and appropriately in their learning experience.
Universities will adapt teaching and assessment to incorporate the ethical use of generative AI and support equal access.
Universities will ensure academic rigour and integrity is upheld.
Universities will work collaboratively to share best practice as the technology and its application in education evolves.

The stance is managed use with integrity, not prohibition.

Publishers: Elsevier and the ICMJE

If your review might become a journal article, publisher rules apply too.

Elsevier's policy for authors, updated in June 2026, accepts that authors may use AI tools to support tasks including synthesizing literature and identifying research gaps, organizing content, and improving language and readability, but says "these tools must never be used as a substitute for human critical thinking, expertise and evaluation." Authors must "disclose the use of AI tools for manuscript preparation in a separate AI declaration statement", with the exception that "Basic checks of grammar, spelling and punctuation do not need a declaration statement." AI tools cannot be listed as authors.

The International Committee of Medical Journal Editors, whose recommendations many biomedical journals follow, requires authors who use AI to "describe, in both the cover letter and the submitted work in the appropriate section if applicable, how they used it". It states that chatbots "should not be listed as authors because they cannot be responsible for the accuracy, integrity, and originality of the work", that humans must "carefully review and edit" AI output because it "can be incorrect, incomplete, or biased", and that "Referencing AI-generated material as the primary source is not acceptable."

Where the policies agree

Principle	Cambridge	Stanford	Toronto SGS	Sydney	Elsevier	ICMJE
AI may support learning and research	Yes	Instructor decides. Default: as help from another person	With supervisor approval	Yes	Yes	Yes
AI use must be acknowledged	Yes	Yes	Yes	Yes	Yes	Yes
Unacknowledged AI content presented as your own is misconduct	Yes	Yes (Honor Code)	Yes	Yes	Not applicable	Not applicable
The human author is responsible for accuracy	Implied	Implied	Stated	Implied	Stated	Stated
Local rules (brief, instructor, supervisor) override the default	Yes	Yes	Yes	Yes	Journal-specific	Journal-specific

What this means for each literature review task

The table applies the principles above to the usual workflow. "Usually fine" means it is consistent with every policy quoted here when no local rule says otherwise. It does not mean your course allows it.

Task	Typical status	Conditions
Finding papers with AI-powered search or recommendation tools	Usually fine	Name the tools in your search description. Verify that every paper exists and says what you think it says. Do not rely on one tool alone
Generating keyword and synonym ideas for database searches	Usually fine	You decide the final search string and record it
Triage: AI summaries to decide which papers to open	Usually fine	Never cite from a summary. Read what you cite
Asking a chatbot to explain a method, statistic or dense passage	Usually fine	Same as asking a tutor. Check the explanation against a textbook if your argument depends on it
Translating a paper written in a language you do not read	Usually fine, disclose	Machine translation can distort technical terms. Say that you relied on a translation
Extracting data into a table (sample sizes, designs, outcomes)	Grey area, disclose	Ask first in assessed coursework. Check every cell against the paper. In a systematic review, describe the tool and the verification in the methods
Grouping sources into themes or suggesting an outline	Grey area	This is part of the analysis you are assessed on. Apply the "another person" test, do your own grouping first, disclose any AI input
Drafting paragraphs of the review	Misconduct under the policies quoted here if unacknowledged and presented as your own, unless the brief permits it	If permitted, disclose precisely. For theses, get supervisor approval in advance
Rewording someone else's text with AI so it looks original	Misconduct	It is plagiarism of the source, whatever tool does the rewording
Grammar, spelling and style checking of text you wrote	Usually fine	Elsevier exempts basic checks from disclosure. Check your own course rules, especially where language itself is assessed
Generating a reference list from a chatbot	Do not	Chatbots can produce references to works that do not exist. See how to check whether a citation is real
Formatting references you have verified	Usually fine	Check the output against the style guide

The line that runs through the table is the one Elsevier draws: tools can support the work, but cannot substitute for critical thinking and evaluation. In a literature review the assessed thinking is the selection, the evaluation and the synthesis. Everything up to the point where you read the papers is logistics, and AI help with logistics is rarely controversial.

Find the rule that applies to you

Rules are layered, and the most specific one wins.

The assignment brief or thesis regulations. Look for a section headed AI, generative AI, permitted tools or academic integrity.
The course or module syllabus.
Department or faculty policy. Some faculties are stricter than their university.
University policy. Search the university site for "generative AI students".
Your supervisor, for a thesis. Get agreement in writing.
The target journal and its publisher, if you intend to publish.

If the first five are silent, ask your instructor by email and keep the reply. Silence is not permission. Stanford's default exists because silence is common.

How to disclose

A good disclosure names the tool, says what you used it for, and says what you did to verify the output. It goes where your institution asks for it: a declaration form, an acknowledgements section, or the methods section.

The two templates below are our own wording, not an official form. Replace the bracketed parts, and delete any sentence that is not true of your work.

For a literature search described in a methods section:

Searches were run in [databases] in [month and year] (full strings in Appendix [letter]). An AI-assisted search tool ([name of tool]) was also used to identify additional papers and suggest search terms. All included sources were retrieved and read in full by the author. No generative AI tool was used to draft or edit the text of this review.

For an acknowledgement in coursework:

I used [name of tool] ([provider, model or version], accessed [dates]) to explain [concept or method] while reading [author and year of the paper], and [name of tool] to check spelling and grammar. I did not use AI tools to generate or reword any of the content of this assignment.

For a journal submission, Elsevier's suggested wording is: "During the preparation of this work, the author(s) used [NAME OF TOOL / SERVICE] in order to [REASON]. After using this tool/service, the author(s) reviewed and edited the content as needed and take(s) full responsibility for the content of the published article."

For how to reference AI output itself in APA, MLA or Chicago, see how to cite AI tools. For the separate question of sources you found through an AI tool, see how to cite sources you found through AI tools.

Keep a process record

An acknowledgement is easy to write if you have kept a log, and hard to reconstruct if you have not. Sydney's student guidance asks for the same thing where AI use is permitted: keep track of how you used generative AI and keep copies of the outputs.

A dated log of AI use: tool, version, what you asked, what you did with the answer. One line per session is enough.
Your saved database searches and result counts.
Reading notes in your own words, with page numbers.
Drafts with version history. Google Docs keeps one automatically, and so does Word for files stored in OneDrive or SharePoint.

Two things policies rarely mention

Confidentiality and copyright. Uploading unpublished manuscripts, interview data or a classmate's draft to a public chatbot may breach confidentiality or research ethics approval. Publisher PDFs obtained through your library are supplied under licence terms, so check those terms, and what the tool does with uploaded files, before you upload anything that is not already public.

Skill. Toronto's warning about skills deserves to be taken at face value. The literature review is where researchers learn to judge evidence and build an argument from it. A review assembled by a tool leaves you unable to defend it in a viva or at a committee meeting, where no tool is available.

Wonders is a search and organizing tool: it helps you find papers, keep them together and trace a claim back to its source, and it does not write your review for you. That puts it on the "logistics" side of the table above rather than the assessed-thinking side, but it does not exempt you from your own rules — Toronto's guidance, for instance, asks for supervisor approval before generative AI is used in any aspect of researching or writing a thesis. Acknowledge the tool and check your brief. A supervisor view of the searches and sources behind a project is in beta for some accounts. We set out the reasoning in Don’t cheat.

Sources

Elsevier. Generative AI policies for journals (updated June 2026). https://www.elsevier.com/about/policies-and-standards/generative-ai-policies-for-journals
International Committee of Medical Journal Editors. Recommendations: Artificial intelligence (AI), AI use by authors. https://www.icmje.org/recommendations/browse/artificial-intelligence/ai-use-by-authors.html
Liang, W., Yuksekgonul, M., Mao, Y., Wu, E., & Zou, J. (2023). GPT detectors are biased against non-native English writers. Patterns, 4(7), 100779. https://doi.org/10.1016/j.patter.2023.100779
Russell Group. (2023). Principles on the use of generative AI tools in education. https://www.russellgroup.ac.uk/policy/policy-briefings/principles-use-generative-ai-tools-education
Stanford University, Office of Community Standards. Generative AI policy guidance (16 February 2023). https://communitystandards.stanford.edu/generative-ai-policy-guidance
University of Cambridge. Plagiarism and academic misconduct: Artificial intelligence (AI). https://www.educationalpolicy.admin.cam.ac.uk/plagiarism-and-academic-misconduct/artificial-intelligence-ai
University of Sydney. Academic integrity: artificial intelligence. https://www.sydney.edu.au/students/academic-integrity/artificial-intelligence.html
University of Sydney. (2024, November 27). University of Sydney AI assessment policy. https://www.sydney.edu.au/news-opinion/news/2024/11/27/university-of-sydney-ai-assessment-policy.html
University of Toronto, School of Graduate Studies. Guidance on the appropriate use of generative artificial intelligence for graduate academic milestones. https://www.sgs.utoronto.ca/about/guidance-on-the-use-of-generative-artificial-intelligence/
Weber-Wulff, D., Anohina-Naumeca, A., Bjelobaba, S., Foltýnek, T., Guerrero-Dib, J., Popoola, O., Šigut, P., & Waddington, L. (2023). Testing of detection tools for AI-generated text. International Journal for Educational Integrity, 19, 26. https://doi.org/10.1007/s40979-023-00146-z