You Think "Real" Is Enough?
In academic writing, the most overlooked step is citation verification.
In the past, the only question was: Does this paper actually exist? The rise of ChatGPT made "hallucinated citations" a hot topic—AI can fabricate references that look perfect but don't exist.
LitSource's first-generation verification addressed this: detecting whether a citation is real, whether the DOI matches, and whether the metadata is correct.
But after serving thousands of biomedical users, we realized: "Real" is just the baseline. "Safe to cite" is where the real danger lies.
A citation can be 100% real and still cause serious problems:
- The paper has been retracted due to fraud
- The paper is only a preprint—never peer-reviewed
- The paper is a narrative review, but you cited it as primary evidence
- The paper is an animal study, but you used it to support a clinical conclusion
- The paper has a published erratum, and the data you cited may have changed
The worst part? Reviewers catch these instantly. You might not even know.
Citation Risk Radar: Your Last Safety Net
This is why we built the Citation Risk Radar.
It goes beyond true/false detection. On top of verifying authenticity, it performs a comprehensive risk scan for every citation and provides clear risk levels with actionable recommendations.
Six Risk Detection Dimensions
| Risk Type | Level | Meaning |
|---|---|---|
| 🔴 Retracted | Critical | Paper withdrawn due to misconduct—must remove |
| 🔴 Expression of Concern | High | Journal has raised formal concerns |
| 🟠 Preprint Only | High | Not peer-reviewed, unsuitable as core evidence |
| 🟠 Animal / In vitro Only | High* | High risk when supporting clinical claims |
| 🟡 Review, Not Primary | Medium | Reviews are not original research |
| 🟡 Erratum / Correction | Medium | Verify whether corrections affect your cited data |
Five-Level Risk Classification
Every citation is classified into a clear risk level:
- 🔴 Critical — e.g., retracted; must remove immediately
- 🟠 High — e.g., preprint, animal-only; not recommended for direct citation
- 🟡 Medium — e.g., review, erratum; requires manual review
- 🟢 Low — no known risks; safe to cite
- 🔵 Info — informational note; no material impact
Complete Risk Report Per Citation
For each reference, the system provides:
- Risk Name — e.g., "Retracted"
- Risk Level — e.g., Critical
- Risk Evidence — e.g., "Listed in Retraction Watch; author had 90+ papers retracted for data fabrication"
- Recommended Action — e.g., "Remove this citation; search for a non-retracted RCT with the same PICO"
Evidence Level Tags
Beyond risk detection, the Risk Radar labels each reference with its position in the evidence pyramid:
- 🔵 Guideline — Clinical practice guideline (highest level)
- 🔵 Meta-Analysis — Pooled analysis of multiple studies
- 🔵 Systematic Review — Structured evidence synthesis
- 🟢 RCT — Randomized controlled trial
- 🟡 Cohort — Cohort study
- 🟡 Case-Control — Case-control study
- 🟠 Animal Study — Animal model research
- 🟠 In vitro — Cell/tissue experiments
- 🟣 Review — Narrative review (not systematic)
- ⚪ Case Report — Individual case description
The system also tags the subject level: Human / Animal / In vitro, helping you instantly determine whether the evidence matches your clinical claim.
Actionable Recommendations
The Risk Radar doesn't just label—it tells you exactly what to do:
| Recommendation | Meaning |
|---|---|
| ✅ Safe to cite | No risk, evidence level matches, use with confidence |
| 📋 Background only | Suitable for Introduction, not for Discussion conclusions |
| 🐀 Mechanism support only | Animal/in vitro results for mechanistic context only |
| ⚠️ Consider replacing | Higher-level evidence is available |
| 📝 Manual review needed | Erratum or partial mismatch; verify manually |
| 🚫 Must remove | Retracted or critically compromised; do not cite |
Real-World Use Cases
📝 Pre-Submission Check
Before hitting "Submit," paste your reference list into Risk Radar for a final scan. Catch retracted or high-risk citations before the reviewers do.
🤖 AI-Generated Reference QA
If you used ChatGPT, Claude, or Kimi to generate references, run them through Risk Radar. AI tools may recommend papers that are retracted or don't exist.
📚 Systematic Review Screening
When writing a systematic review with 100+ references, Risk Radar quickly filters out reviews, animal studies, and preprints that shouldn't serve as core evidence.
📄 Grant and Ethics Writing
Citations in grant applications and IRB materials must be bulletproof. Risk Radar ensures every reference is verified, risk-free, and traceable.
✍️ Revision Response
When reviewers request additional citations, new references need the same level of scrutiny. Risk Radar prevents introducing new citation risks during revision.
Why Biomedical Only
Biomedical research has a far lower tolerance for citation errors than other fields. A retracted paper in a clinical decision chain can affect treatment protocols and patient safety.
Citation Risk Radar is purpose-built for biomedical contexts:
- Leverages PubMed PublicationType and MeSH terms for study classification
- Uses Crossref and Retraction Watch for retraction and erratum detection
- Explicitly distinguishes human studies from animal/in vitro research
- Risk levels align with the clinical evidence pyramid
We don't do generic, cross-discipline checking. We do deep biomedical evidence risk control.
In One Sentence
Citation Risk Radar: It doesn't just tell you if a reference is real—it tells you if it's safe to cite.
Three minutes before submission. One complete risk check for your entire reference list.