# Fraud detection and response quality

Canonical page: https://www.questionpunk.com/support/fraud-detection

Automatic detection of suspicious responses with quality classification.

Time: 5-10 minutes to configure

Level: Intermediate

Audience: Researchers, Analysts

## Steps

### 1. Automatic classification

Every response is automatically analyzed and assigned a suspicious level: **low** (likely genuine), **medium** (some concerns), or **high** (likely suspicious).

### 2. Filter responses

In the Results tab, use the filter dropdown to show only low-suspicious responses, or review high-suspicious ones manually.

### 3. Review signals

Each response shows its session details: IP address, location, device info, time spent, and events. Use this to verify response quality.

## Details

Fraud detection runs automatically on every response. The system checks typing behavior (speed, paste/copy use, backspacing), response timing, and straight-lining patterns on rating scales.

Suspicious level is included in CSV exports, so you can filter at the analysis stage. Use the Results tab filters to review flagged responses before exporting.

Use [respondent codes](https://www.questionpunk.com/support/audiences) to tag flagged responses with custom labels (e.g., "Needs review", "Confirmed valid") for organized follow-up.

For studies using panel providers like Prolific, fraud detection helps identify participants who may not be engaging genuinely with your survey.

## Related articles

- [Synthetic testing with AI personas](https://www.questionpunk.com/support/synthetic-testing.md)
- [Design for better responses](https://www.questionpunk.com/support/survey-design-best-practices.md)
- [Exporting results](https://www.questionpunk.com/support/data-exports-exporting-results.md)