ChatGPT and other large language models tend to invent information, so-called hallucinations. This is particularly evident with structured data such as references and bibliographies. Recent studies have shown that above 25% of the generated citations contained errors or were completely fabricated!
Our online tool takes a literature list with one citation per line and will check them for existence and correctness. Paste yours below or test with our sample bibliography.
This interpretation only works if the document had regular saving points (i.e., auto-save was active) while it was written; otherwise, copying the text into a fresh .docx file removes this metadata, making the analysis meaningless.
Got your word file stores in MS Drive or Google Drive? Try our Version History Tool to track the evolution of your documents
To get a better analysis, use Mentafy from the start of your project. Read More
Work with Mentafy from the beginning of your writing, because
We will generate a concise report with indicative thresholds, which allows you to judge authenticity within a few minutes and generate a concise report with indicative thresholds, which allows you to judge authenticity within a few minutes.
We provide the authors with writing support from the first steps of finding a topic until the final submission. Most importantly, real-time feedback for the students, encouraging them to work diligently and honestly.
We track document changes reliably (Note: Even if you move text around within a document, we keep the information, how it came in originally) We are not keyloggers, and we respect our users privacy.
FAQ
Plagiarism compares text overlap; AI detection estimates generation patterns. Our Reference Verifier checks whether citations and references are real, matched, and consistent, touching a different integrity layer.
Accuracy is high for well-formatted, standard styles. False flags can occur with non-standard formats, incomplete metadata, or very obscure sources. The UI makes each flag auditable, so you can quickly confirm or correct.
No. It provides evidence to inform human judgment. Final decisions remain with educators and institutions.
Yes, partially. We match using author/title/year/venue where available. Non-DOI items may need manual confirmation in the UI.
Yes, if they contain standard scholarly references. Legal/Bluebook-style footnotes and id./ibid. chains aren’t fully supported yet.
Groups like (Smith 2021; Jones 2020; Li 2019) are parsed into individual pairs, each checked against the bibliography and external registries.
If a DOI is invalid or a referenced source does not exist in public registries, we flag it. This does not prove AI use right away. If formatting flaws are not the reason and this happens a lot in the paper, it's a strong risk signal.
It is designed for academic-level work (theses, seminar papers, research reports) that follows strict citation standards. For casual assignments with loose formatting, expect more manual review.
Reference Verifier sits inside the Mentafy Full Experience:
Writing Pattern Recognition: originality & AI-use alerts
Research Recorder: captures and organizes sources
Authorship Report: defensible authorship trace
Plagiarism Checker & AI Recognition: complementary signals
For best results on verifying a paper's authenticity we recommend using the Full Mentafy Suite which grasps a detailed overview of the whole writing project from first word to submission.
Best results with Latin-alphabet languages and standard transliteration. Non-Latin scripts can work when core fields (author/year) are in Latin script.
Yes. Mentafy is designed from the ground with GDPR/CCPA principles in mind (data minimization, purpose limitation, user controls). Learn more >