Deepfake Detector for Images
Screen a still image for signs of AI generation or editing, then choose the next check for the manipulation you suspect.
Check a photo
What kind of manipulation are you looking for?
A fully generated portrait, a face swap, a retouched photograph and a real image with a false caption are different problems. This release provides general AI-image generation and editing signals; it is not a dedicated face-identity or liveness verification service.
| What you suspect | What the image check can tell you | What to check next |
|---|---|---|
| Entire image generated | The detector can flag signals consistent with AI generation. | Look for an original file or creation disclosure. |
| Face swap | General editing signals may be present. The result cannot establish that a face was replaced. | Compare with trusted images or the original sequence. |
| Ordinary photo edit | An edited category may be returned. It does not explain the purpose of the edit. | Ask for the uncropped original or the edit history. |
| Real image, false caption | The file can look authentic even when the claim is wrong. | Find the earliest publication and verify the date, place and event. |
How to review suspected deepfake imagery
Use the result as an initial screen. For a suspected face swap, compare the media with a trusted original, inspect the surrounding video or photo sequence and use a specialist identity-analysis service when the decision carries serious consequences.
Preserve the original
Keep the highest-quality file and its surrounding post or message. Screenshots may remove metadata and visual detail.
Separate the manipulation type
Ask whether the whole image may be generated, a face may be replaced, the photo may be edited, or the caption may simply be false.
Escalate serious cases
For fraud, employment, legal or safety decisions, obtain specialist review and corroborating evidence instead of relying on one automated score.
- Can this tool confirm a face swap?
- No. It can screen the image for general generation or editing signals, but it does not perform dedicated face-swap attribution or identity verification.
- Does a likely-real result prove the person is genuine?
- No. Likely real describes the provider classification for the image. It does not verify who the person is, whether the account belongs to them or whether the caption is true.
- Should I use the score for legal or financial decisions?
- No. Automated image detection can be wrong. Consequential cases require source records, additional forensic methods and qualified human review.