Before your first check
Know what you’ll get.
A useful report tells you what was checked, what the tool found, and what it cannot establish. These examples let you explore that without uploading anything.
Image: a completed operational check
We screened CouldItBe’s own public social-preview graphic on November 4, 2026. The request completed through the live site and Sightengine. This single test verifies the workflow; it is not an accuracy benchmark or a representative photo test.
- Provider / model
- Sightengine / genai
- AI-generation signal
- 2%
- Model version
- Not supplied by the provider
- Completed
- November 4, 2026, 22:30 UTC
How to read it: this input produced a low AI-generation signal. It does not certify the image’s origin, its publishing rights, or the accuracy of its contents. The percentage is not the share of pixels generated by AI.
Download sample image report (JSON)View the public graphic used
The download is marked as an example and contains the actual provider result. It contains no image bytes. Check your own image →
Text: understanding an inconclusive result
Illustrative example · Not a real scan
AI screening: Inconclusive.The available screening result does not support a confident classification. Keep this uncertainty in the record; do not convert it into “human-written.”
A completed text report can include the classification, provider confidence, model version when supplied, completion time and separate local observations. Our example leaves unmeasured fields explicitly unmeasured and does not invent a provider score.
Download illustrative text report (TXT)
Your actual text tool offers a readable report, JSON and print/save PDF controls after analysis. Source writing is excluded from its downloads. Review your own writing →
Three states worth keeping separate
- Completed: a provider returned an assessment. It still needs context and can be wrong.
- Inconclusive: the screening does not support a confident classification. More context may help; a verdict is not guaranteed.
- Not completed: no usable detector result exists—for example, because you chose local review, cancelled, or the service was unavailable. Local observations do not replace a detector result.
For video, reports describe the sampled frames, not the whole video. Batch results stay separate for each file; there is no combined authenticity score.
Before sharing a report
Read the file you downloaded. Batch reports include the filenames you supplied, even though they exclude source file contents. Rename sensitive files before adding them, or remove identifying filenames from the downloaded report before sharing. Keep your original content and editorial notes securely yourself.
Downloads are not signed forensic certificates. Record the relevant source and context separately, check citations and permissions, and never make an accusation based only on a detector score. Read the privacy policy and result guide.