Most advice about fabricated references stops at "search for it, and if it's not there, it's fake." That's too blunt. Plenty of real sources never show up in Google Scholar or Crossref, and throwing them out is its own kind of error.
So here's the other half of the problem: what to do when a reference returns nothing.
First, rule out the obvious
Make sure you searched the exact title in quotes in more than one place: Google Scholar, then Crossref Search or OpenAlex. A topic search doesn't count, because it will always find something relevant, and that's exactly how an invented reference starts to feel real.
Then look where real sources hide
- Older books and chapters. Try WorldCat or Open Library with the ISBN or title. Be careful with chapters, though: they're also the format AI fabricates most.
- Local and national journals that never registered DOIs. Go to the journal's own archive.
- Theses, reports and conference posters. Check the repository of the institution that published them.
- Non-English sources. Search the title in its original language, not a translation.
- Web sources. If the link is dead, try the Wayback Machine. A page that was never archived anywhere is a red flag.
If it still isn't there
Don't cite it. Ask whoever gave you the reference for a copy. If they can't produce one, that's your answer.
Where a tool helps
Doing this by hand for one reference is fine. For a full bibliography it's slow, so I built Fake Citation Detector, a free Chrome and Edge extension. It searches six databases (Crossref, OpenAlex, DataCite, PubMed, Open Library and Semantic Scholar) and sorts each reference into confirmed, wrong details, not found or web source, checking dead links against the Internet Archive along the way. Crucially, it treats "not confirmed" as "look closer", never as a verdict.
The complete five-step method to check if a citation is real, with a red-flag table and FAQ: How to Check If a Citation Is Real: A 5-Minute Method