Trust & Transparency

Sample Detection Reports

These examples show what different scan results look like and how to interpret them. All examples below are from internal test data — they are not real user scans.

Report examples

Likely AI — High confidence

Verdict: Likely AI

Confidence: High

A short clip showing a person speaking with unnatural lip-sync timing and inconsistent facial lighting across frames.

Signals detected:

Lip-sync mismatch at 0:03–0:07
Facial landmark instability at 0:12
Lighting inconsistency on forehead vs. cheek area

Source: Synthetic test clip created with an open-source text-to-video model for internal testing.

This is an internal test example, not a real user scan. It illustrates the type of result a high-confidence AI detection produces.

Unclear — Low confidence

Verdict: Unclear

Confidence: Low

A YouTube reupload of what appears to be a real nature documentary clip, heavily compressed by the platform.

Signals detected:

Mild temporal inconsistency between frames 0:08–0:11
Platform compression warning — reduced signal quality

Source: Public domain nature footage, re-encoded through a social platform simulation to test compression effects.

Low-confidence results are common on compressed sources. This does not mean the video is AI-generated — it means there was not enough signal for a strong verdict.

Likely Real — Medium confidence

Verdict: Likely Real

Confidence: Medium

An original MP4 of a street interview recorded on a consumer smartphone, with no post-processing.

Signals detected:

No significant AI-generation signals detected
Consistent lighting and motion across all frames

Source: Internal test recording made with a standard smartphone camera.

A low score does not prove a video is real. It means the available evidence did not suggest AI generation. High-quality AI output can also produce low scores.

Likely AI — Medium confidence (face swap)

Verdict: Likely AI

Confidence: Medium

A clip where one person's face appears to have been swapped onto another person's body, with subtle boundary artifacts.

Signals detected:

Skin texture inconsistency at jawline boundary
Eye blink asymmetry at 0:05
Shadow mismatch between face and neck

Source: Synthetic test clip using open-source face-swap software for internal testing.

Face-swap detection focuses on facial region analysis. Results are most reliable when the face occupies a reasonable portion of the frame.

About these examples

All examples shown are from internal test data created for demonstration purposes. They are not real user scans.
Test clips were created using open-source AI generation tools and standard consumer cameras.
Real scan results will vary based on video quality, compression, length, content, and the specific AI model used.
These examples illustrate the report structure and signal types — they are not benchmarks of detection accuracy.
For details on how detection works, see Methodology.
For known limitations, see Accuracy & Limitations.