Short Description:
A 5-minute, method-first activity: students verify three references from their own draft bibliography before submitting - resolving DOIs and searching exact titles - then classify each as found-and-matches, found-but-differs, or not-found, and correct entries from the real record. Motivated by the rise of fabricated ("hallucinated") citations that look perfectly formatted yet point at nothing. Teaching point: "not found" does not automatically mean "fabricated".
Learning Outcomes:
By the end of this activity, students will be able to: (1) verify a reference by resolving its DOI at doi.org or searching the exact title in a scholarly database; (2) classify each reference as found-and-matches, found-but-differs, or not-found; (3) correct bibliographic entries using the real source record; (4) explain why a confidently formatted reference that resolves to nothing anywhere carries the signature of a machine-invented citation, while "not found" alone does not prove fabrication.
Course Context (e.g. how it was implemented or integrated):
Designed as a 5-minute opener for any course whose assignments include references: information literacy one-shots, first-year writing, research methods, or graduate seminars. Best run shortly before a submission deadline, when students have a draft bibliography in hand. No preparation needed; works in person or online. Materials: the student's own draft bibliography and a browser.
Additional Instructor Resources (e.g. in-class activities, worksheets, scaffolding applications, supplemental modules, further readings, etc.):
Student-facing script: (1) Pick three references from your bibliography: one from a source you actually read, one that was suggested or produced by any writing tool, and the one you are least sure about. (2) Try to reach each work. If it has a DOI, paste it at https://doi.org. If not, search the exact title in quotes in Google Scholar or your library's discovery search. (3) Classify each reference: Found-and-matches (authors, year, and venue agree with your entry); Found-but-differs (a real work exists, yet your metadata disagrees); Not-found (the DOI resolves to nothing and no database returns the title). (4) Act on it: for found-but-differs, correct your entry from the real record; for not-found, try one more route (a second database, the journal's own site) before drawing any conclusion.
Teaching point: "not found" does not automatically mean "fabricated" - older books, regional journals, and grey literature are genuinely under-indexed. But a reference that is confidently formatted, carries a plausible identifier, and resolves to nothing anywhere is the classic signature of a machine-invented citation - treat it as unusable until the real source is located.
Optional discussion statistic: in one recent 30-day window, 47,098 references were run through a public reference checker and 27.2% triggered a fabrication signature - a conservative flag rate, not a validated prevalence estimate. See the relevant links for the write-up and the open CC BY 4.0 dataset.
This activity is adapted from the CC BY 4.0 classroom resource by CiteMe (citeme.app). Adapt freely with attribution. A Portuguese (PT-BR) version is available from the source.
Potential Pitfalls and Teaching Tips:
