AI in academic research
Two questions dominate: am I allowed to use this, and can I trust what it gives me. These pages answer both with dated policies from the publishers and funders themselves, and with published measurements rather than opinion.
Updated
The guidance on AI in research is scattered across publisher policy pages, funder announcements, university handbooks, and departmental rules that contradict each other. Meanwhile the practical question, whether a given output can be trusted, is answered mostly by anecdote.
These pages take both seriously. Where a rule exists, it is quoted, dated, and linked to the organization that published it, because these policies changed several times between 2023 and 2026 and a page without a date is worthless. Where a measurement exists, the study is named with its numbers, its model, and its year, because there is no such thing as a single fabrication rate.
We build an AI research tool, so treat our position as declared rather than neutral. The declaration cuts in an uncomfortable direction: the strongest finding in this literature is that a language model asked to recall citations from memory produces references that do not exist at rates that make the output unusable for scholarship. That is why the honest advice on several of these pages is to verify every reference yourself, whatever tool produced it.
Why does ChatGPT make up references?
Why language models produce citations to papers that do not exist, what published studies measured about how often it happens, and how to check a reference fast.
Read the answer →Can I use AI for a literature review?
Whether AI can be used for a literature review, which parts of the process it genuinely helps with, and which parts have to stay human to survive examination.
Read the answer →Is using ChatGPT for academic writing plagiarism?
Whether using ChatGPT in academic writing counts as plagiarism, where the line between editing and generating falls, and what university policies typically say.
Read the answer →Do I need to disclose AI use in a research paper?
When AI use must be disclosed in a research paper, what the major publishers require and where the statement goes, and which uses are exempt from declaration.
Read the answer →Can an AI tool be listed as an author on a paper?
Whether an AI tool can be named as an author on a research paper, what COPE and ICMJE say about it, and how to credit AI assistance properly instead.
Read the answer →Are AI detectors reliable?
What published studies found about AI text detector accuracy, why detectors flag non-native English writers more often, and what to do if your writing is flagged.
Read the answer →How do I cite ChatGPT in APA 7?
How to cite ChatGPT and other AI tools in APA 7, MLA 9, and Chicago style, and whether a chat should be cited at all or described in the methods instead.
Read the answer →How do I check whether AI-generated citations are real?
How to verify AI-generated citations in a few minutes, what a fabricated reference usually looks like, and how to confirm the source says what was claimed.
Read the answer →Can I trust AI summaries of research papers?
How reliable AI summaries of research papers are, what summarizers systematically get wrong, and how to check a summary in under two minutes.
Read the answer →Can I use AI in a PhD thesis?
Whether AI can be used in a doctoral thesis, what university policies typically permit and require, and how to raise the question with your supervisor.
Read the answer →Can I use AI for screening in a systematic review?
Whether AI can be used to screen studies in a systematic review, what reporting guidelines require you to declare, and how to validate an automated screening step.
Read the answer →Why does an AI answer better when it reads my own PDFs?
Why an AI gives better answers when it works from documents you supplied, what retrieval is and why it reduces fabrication, and whether it removes errors entirely.
Read the answer →Can I use AI to help write a peer review?
Whether AI can be used to help write a peer review, what uploading an unpublished manuscript to a chatbot risks, and what publishers and funders currently allow.
Read the answer →