ReferenceVerify
Go ProSign in TR EN

Verify citations

Give us your text and bibliography; we check whether each citation is really supported by its source.

This tool needs a free account. Your first day includes 60 citations (up to 5 runs); after that you can verify 20 citations a day. No card required. Pro: 250 a month, Max: 1,000 a month.

Have the full text of closed-access sources? (optional: PDF or Word)

Upload the full text of the sources (PDF, Word .docx or .txt). The text is extracted IN YOUR BROWSER and auto-matched to the right reference, so subscription-only sources get verified with full text too. For PDFs the evidence page is shown as well.

Paste text manually (advanced)
Sign in / Sign up
How to use this tool

One step beyond the reference check: even when a source is real, does the sentence citing it actually match what the source says? Each citation is marked Full / Partial / Unsupported / Could-not-verify.

  1. Open “Verify citations” and paste your text (or upload the file) together with its bibliography.
  2. If you have the full text of closed-access sources, upload it (PDF, Word .docx, .rtf or .txt); the text is extracted IN YOUR BROWSER and auto-matched to the right reference, so subscription-only sources get verified with full text too.
  3. Glance at the matching table: it shows which file was linked to which reference and why (DOI, or title/author). Fix any wrong link with one click.
  4. Start the run. The bibliography is verified first, then your text is split into claims and each claim is compared, by meaning, with the cited source, against its full text when it can be accessed, otherwise against its abstract. The comparison is made by a local AI model on our own server.
  5. Watch progress: a time estimate is shown; expect 1–5 minutes depending on the number of sources.
  6. Read the report: per-citation verdicts with evidence, plus a corrected bibliography. When the evidence comes from a file you uploaded it is labelled as such, and for PDFs the page is shown too (“your file · p. 7”), so you can confirm it in seconds. Download as Word or CSV, or share it via link.
Full guide & PDF →

Analysis usually takes 1–5 minutes and runs in this tab. Don't close the page or navigate within the site before it finishes.

Where references are checked, and how your text is protected

Sources are verified in international and national scholarly databases; content is read from open-access full texts and abstract records. Your text is never sent to outside AI services or any other third party, never used for training and never passed to plagiarism databases. A report is kept in your account only when “save to my account” is ticked.

Sample result: the four verdicts

The tool compares each cited sentence in your paragraph with the abstract or full text of the source it cites, and returns one of four verdicts with its reasoning. The cards below are an example in the output format.

Full matchSource 1from full text

The Transformer architecture relies solely on attention, without any recurrent layers [1].

Reasoning: The source states that the model is based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. The claim matches the source.

Partial matchSource 1from full text

The Transformer outperforms all previous models on every natural language processing task [1].

Reasoning: The source shows superiority on two machine translation tasks (English–German, English–French) and reports good results on constituency parsing. The generalisation to “every task” is not in the source.

UnsupportedSource 2from full text

Like GPT, BERT is pre-trained by reading text only from left to right [2].

Reasoning: The source says the opposite: BERT is pre-trained bidirectionally with a masked language model, using left and right context together, exactly what sets it apart from unidirectional models.

Could not verifySource 3

Scientific revolutions proceed through paradigm shifts rather than cumulative progress [3].

Reasoning: The reference was confirmed in library catalogues, but no open-access full text of the book exists, so the claim could not be compared with its content. Paste the relevant pages and it is re-checked against the full text.

  1. Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, Ł., & Polosukhin, I. (2017). Attention is all you need. Advances in Neural Information Processing Systems, 30. https://doi.org/10.48550/arXiv.1706.03762
  2. Devlin, J., Chang, M.-W., Lee, K., & Toutanova, K. (2019). BERT: Pre-training of deep bidirectional transformers for language understanding. Proceedings of NAACL-HLT 2019, 4171–4186. https://doi.org/10.18653/v1/N19-1423
  3. Kuhn, T. S. (1962). The structure of scientific revolutions. University of Chicago Press.