Citation Risk builds verification infrastructure for the gap between fluent AI output and reliable evidence.
Start with your references. Check whether citations exist, whether key bibliographic details match public records, and where human review is still needed.
A 2026 Lancet study of biomedical papers found the rate of papers containing fabricated references rose from roughly 1 in 2,828 in 2023 to 1 in 277 in early 2026.
That is why citation checking needs to happen before submission.
A source can be real. A DOI can resolve. A citation can be perfectly formatted. And the claim attached to it can still drift beyond what the original evidence actually supports.
Citation Risk starts with citation identity and metadata integrity. Our broader work focuses on preserving the trail between claims, sources and evidence.
Each layer should remain traceable.
Most citation verification starts with bibliographic metadata. Citation Risk goes further by inspecting supported source URLs and comparing the scholarly identity declared on the page with the citation you were given.
We check the citation. We check the identifier. And where possible, we check the source page itself.
Source page inspection
For supported URLs
URL inspection does not automatically prove that the cited source supports the surrounding scientific claim.
Citation Risk checks academic-style references against public bibliographic records and available sources, so you can see what matches, what conflicts, and what still needs human review.
Trust boundary: Citation Risk verifies citation identity and bibliographic metadata. A verified citation does not automatically mean that the source supports every claim attached to it.
Each transformation can improve accessibility, but can also remove context, uncertainty, provenance, and evidential boundaries.
Generation is becoming cheap. Verification remains expensive.
The goal of Citation Risk is to help close that verification gap.
Read: Welcome to the Age of Processed InformationCitation verification is the first layer. We are also developing Structured Intelligence, a governed evidence architecture designed to preserve provenance, distinguish evidence from interpretation, and make the relationship between claims and source material more inspectable.
The goal is not to make AI sound more certain.
It is to make it possible to verify why an answer should be trusted.
Structured Intelligence is ongoing research and development work, not a fully available commercial product.
The principles that guide how Citation Risk is built.
Prefer evidence that can be inspected.
Use structural verification before probabilistic inference.
Do not turn incomplete evidence into confident conclusions.
If evidence does not match, say so.
Automation should surface uncertainty, not hide it.
Verification before trust.
Check messy reference lists, AI-assisted drafts, and suspicious bibliographies before delivery. Use clear verdicts to focus review time where it matters.
Students often bring incomplete, AI-generated, or unreliable references. Citation Risk gives you a quick first-pass audit so you can guide corrections with confidence.
Paste suspicious citations or reference lists and catch fake, mismatched, incomplete, or unverifiable sources before they become your problem.
No full paper upload required.
Add citations, reference lists, or citation-containing text.
Citation Risk checks public bibliographic records and compares available metadata.
Each citation receives a clear verdict, with an audit trail showing what matched, conflicted, or needs review.
Use the results to correct weak, wrong, or unverifiable citations before the document leaves your desk.
The source appears to exist and key metadata matches.
A likely source was found, but supplied details do not match the public record.
No reliable matching source was found. This does not prove the source does not exist anywhere.
Possible match or ambiguity needs human checking.
Not enough citation data was supplied for reliable verification.
A lookup source was unavailable or limited, so the result should be reviewed.
Trust boundary: Citation Risk checks citation existence and bibliographic metadata accuracy. It does not prove that a source supports a claim.
Citation Risk does not try to guess who wrote the text. It does not format citations into APA or MLA. It checks whether references appear to exist and whether their metadata matches the public record.
A formatted citation can still be fake.
A related source is not a verified citation.
A real source can still be a wrong citation.
Access is currently open. Check up to 50 references for free.
After entering your email, you'll be taken directly to the checker.
Currently being tested with real bibliographies across academic, editorial, and AI-assisted writing workflows.
No. Citation Risk checks citation existence and bibliographic metadata accuracy. It does not assess whether the source supports the surrounding argument.
Verified: a matching record was found and metadata appears consistent. Mismatch: a source may exist but details conflict with the record. Not Found: no reliable match was found. Review: a possible match needs manual confirmation. Incomplete: not enough data to verify. Lookup Limited: a lookup source was unavailable, so review the result manually.
Where supported, Citation Risk can inspect an author-supplied or cited URL and extract page-declared scholarly identity information from the source page itself, then compare it with the citation. This does not prove that the cited source supports the surrounding scientific claim.
No. Citation Risk works from pasted citations, reference lists, or citation-containing text.
No. Citation Risk is not a citation generator or formatter. It checks citation risk before users trust or polish the reference.