ChatGPT hallucinated my references — what now?
Fake citations from ChatGPT are common. Learn how to fix your reference list, verify sources, and avoid academic integrity issues.
If ChatGPT gave you a reference list that looks perfect but papers cannot be found, you are not alone — hallucinated citations are one of the most common AI research failures in 2024–2026.
Do not submit the list. Fabricated references — even if accidental — can be treated as academic misconduct regardless of whether you knew the citations were fake.
Audit every entry: search the exact title in Google Scholar, PubMed, or your library. If zero results appear repeatedly, assume the citation is invented.
Rebuild from scratch using academic search. Start with your research question, find real papers, save them to a project workspace, and read at least abstracts before citing.
Talk to your tutor if you already submitted a draft with fake references. Early disclosure and a plan to correct may be viewed more favourably than discovery at marking.
Change your workflow going forward: never paste AI bibliography suggestions directly into assignments. Generate references only from sources you have opened and verified.
Consider this a lesson in why general language models are not reference databases — they predict text, not bibliographic records.
Where Scholise helps
Switch to Scholise for source discovery and reference export — every entry comes from a paper you saved and can verify, not from model memory.
Related guides
Related comparisons
Research with verified sources
Scholise — evidence-first research for students.
Start researching free →