Citation Claim Checker
Paste a sentence with a citation. We check whether the paper actually says what you think it says.
We don’t train AI on your data. Review data handling before pasting sensitive text.
Evidence reviewed · Verify v1.1 · Data handling · Methodology and sources
Paste your claim
A sentence from your manuscript that cites a paper (author-year or numbered)
We resolve the source
DOI-first lookup or a candidate match across three scholarly indexes
Inspect the verdict
Five bounded outcomes with the matched paper and an explanation
Evidence path
What Verify actually checks
1. Resolve one paper
A DOI is used directly. Otherwise, author, year, title words, and claim terms are used to rank candidate records from Semantic Scholar, Crossref, and PubMed.
2. Require source text
A verdict requires a substantive abstract. When the abstract is missing or too short and a DOI is available, Verify tries an eligible open-access full-text source.
3. Compare one claim
The model receives your sentence and the available source text. It must return one of five defined outcomes, not an assessment of the manuscript as a whole.
4. Surface a retraction signal
When the matched paper has a DOI, Verify checks Crossref retraction metadata. No flag is not proof that a paper is current or otherwise reliable.
Primary service documentation
Limits to use when interpreting a verdict
- Candidate matching can select the wrong paper, especially for common surnames. A DOI or full reference is safer than author and year alone.
- Open-access full text is attempted only when the abstract is missing or under the source-text threshold. Tables, figures, and supplementary files are not evaluated.
- The retraction signal depends on metadata available for the matched DOI. Use the Reference Checker or the publisher record for a broader integrity review.
- The verdict is model-generated and has not been presented here as a calibrated accuracy estimate or a substitute for reading the paper.
Calibrated from early work with 35+ CNS-experienced reviewers and senior scientists. Their reviewer language shaped how Manusights flags unsupported claims, weak evidence, journal-fit risk, and likely reviewer objections. Logos indicate researcher affiliation, publication context, or calibration-network context, not endorsement.
Check your entire manuscript
Citation-risk checks, methodology assessment, journal fit analysis, and a prioritized fix list. Delivered in about 30 minutes.
Free preview first. Pay only if the full report is useful.
We don’t train AI on your data. Inputs use short-lived result caching.
Want the data sources behind every verdict? Read the full methodology · search order, source-text gates, retraction checks, refusal cases, and known limitations.
Checking whether references are retracted, not just miscited? Run the Reference Checker · paste a bibliography, get per-reference flags for retractions, expressions of concern, and hijacked-journal clones.