Fraud detection is available as part of Nova’s scoring pipeline. Contact your account manager to enable it for your organization.
How It Works
Nova scores the candidate
When a new application comes in, Nova scores it against your configured criteria as usual.
High-scoring candidates are analyzed for fraud
If the candidate scores at or above your fraud detection threshold (default: 7/10), Nova runs a separate fraud analysis. This analysis combines resume text analysis with multiple independent external validation signals.
What Nova Analyzes
Nova evaluates resumes across multiple independent dimensions. When concerns from different dimensions converge, the fraud probability increases. When only surface-level issues are present, risk stays low.| Dimension | What it checks |
|---|---|
| Timeline integrity | Overlapping dates, impossible sequences, future-dated roles |
| Career progression | Seniority jumps that skip expected intermediate levels |
| Skill credibility | Implausibly broad technology lists spanning unrelated domains, keyword stuffing |
| Quantitative claims | Repeated large impact numbers without baselines or context |
| Document metadata | Suspicious PDF creation tools and document assembly artifacts |
| Hidden content detection | Invisible text via tiny font sizes, keyword stuffing, overlapping duplicate text layers |
| Email verification | Email deliverability, account age, and breach history. Contact data is sourced from ATS structured fields first, falling back to resume text extraction. |
| LinkedIn validation | Profile completeness, activity history, identity coherence, and timeline consistency. When available, Nova can reason over structured LinkedIn profile context in addition to summary signals. |
| Cross-candidate detection | The same email, phone number, or LinkedIn profile appearing across multiple different applications within your company. Reapplications from the same candidate are excluded. |
| AI generation patterns | Resumes that read like uniform templates with zero authentic workplace detail |
Nova distinguishes between normal resume polish (which is standard career advice) and pervasive patterns that suggest fabrication. A few metrics-driven bullet points are not suspicious. An entire resume that reads like a template with no project names, team names, or specific workplace detail is a different story.
Reading the Risk Assessment
Each fraud analysis produces a short note in your ATS designed for fast review:- Risk: Low, Medium, or High
- Takeaway: One-line summary of the main concern or all-clear
- Top reasons (up to 3): Key evidence from the resume or automated checks
- Reason wording: Direct factual statements in plain English, not internal process language
- Reassuring signs (optional): Brief indicators that reduce concern when risk is low
Risk levels explained
| Risk level | What it means | Recommended action |
|---|---|---|
| Low | No meaningful concerns found | Proceed normally |
| Medium | Some structural concerns or signal convergence | Review the top reasons and verify claims during screening |
| High | Strong evidence of fabrication from multiple independent sources | Prioritize verification before advancing |
How it looks in your ATS
- Ashby
- Lever
- Pinpoint
- Greenhouse
- Teamtailor
Benchmark and Accuracy
We maintain an internal benchmark of hundreds of real resumes sourced from production hiring pipelines, including confirmed fraudulent applications and verified legitimate ones. Every change to the fraud detection system is validated against this benchmark before reaching production.Key metrics at the recommended operating point
| Metric | Value | What it means |
|---|---|---|
| Precision | Over 93% | When Nova flags a resume, it is correct over 93% of the time |
| False alarm rate | Under 4% | Fewer than 1 in 25 legitimate resumes are incorrectly flagged |
| High-confidence precision | 100% | At the highest confidence level, Nova has zero false positives across the full benchmark |
Why only high-scoring candidates are checked
Nova only runs fraud analysis on candidates who score at or above your configured threshold (default: 7/10). There are two reasons for this:- These are the candidates your recruiters will actually review. A candidate who scores 3/10 is unlikely to progress regardless of whether their resume is genuine. Fraud detection adds the most value on candidates who would otherwise move forward in your pipeline.
- Fraudulent resumes tend to score high. AI-generated and fabricated resumes are often designed to match job requirements closely, which means they score well against your criteria. The candidates most likely to be fraudulent are concentrated in the high-scoring group.
Configuring Fraud Detection
Open the Fraud Detection page in your Nova dashboard and click the settings icon (gear) in the top right corner.Settings
| Setting | Description | Default |
|---|---|---|
| Enable fraud detection | Toggle fraud analysis on or off | Disabled |
| Score threshold | Minimum Nova score (0-10) required to trigger fraud analysis | 7 |
Fraud Detection Dashboard
When fraud detection is enabled, you get a dedicated dashboard showing all screened applications with their risk levels.What you will see
| Section | What it shows |
|---|---|
| Flagged rate badge | Percentage of screened applications rated medium or high risk |
| Summary cards | Total assessed, plus counts for high, medium, and low risk. Click a card to filter the table to that risk level. |
| Risk trend chart | How the percentage of flagged applications is changing over time. A rising line means more applications are being flagged. |
| Assessment volume chart | Number of applications screened per period, broken down by risk level |
| Assessment table | Full list of screened applications with risk level, verdict, and fraud probability. Click any row to see the detailed analysis. |
Risk assessment detail
Clicking a candidate row opens the full risk assessment, showing the risk level, probability, summary, red flags with severity and evidence, and metadata such as assessment date and ATS note status.
Frequently Asked Questions
What happens if fraud detection fails?
What happens if fraud detection fails?
Fraud detection is designed to fail gracefully. If the analysis encounters an error, the candidate’s score and ATS note are unaffected. The scoring result is always delivered regardless of whether fraud detection succeeds.
What should I do when a candidate is flagged?
What should I do when a candidate is flagged?
Use the top reasons in the risk assessment as your interview checklist. Most concerns can be resolved by asking the candidate to walk through specific claims on their resume. A medium risk flag means “verify before deciding,” not “reject the candidate.”
Can corporate email addresses trigger false positives?
Can corporate email addresses trigger false positives?
Some corporate mail servers return “undeliverable” for external verification attempts as a security measure. Nova accounts for this in its scoring, but if you see email-related flags for candidates using corporate addresses, this is likely the cause. Review the other signals in the assessment.