Free AI Face Swap Detector
Has this face been swapped by AI? Check photos and videos and review the evidence in seconds.
Free, no signup, uploads are not retained, with clear uncertainty.
What do you want to check?
What Is an AI Face Swap?
A face swap is a deepfake where a real photo or video is kept, but the face is replaced with someone else's. Modern tools adapt the angle, lighting and expressions of the new face so the result can be extremely convincing. What used to require hours of training with DeepFaceLab now happens in one click in commercial apps — and real-time tools can even swap a face live inside a Zoom, Teams or Google Meet call.
Face swaps drive some of today's most damaging fraud: fake celebrity investment endorsements, political ads impersonating candidates, romance scams built on stolen faces, and job candidates who are not who they appear to be. Identity-fraud reports tracked face-swap attacks growing over 300% year on year, with detection by eye increasingly unreliable — studies put human accuracy on high-quality video deepfakes below 25%.
How to tell if a photo is face swapped
A face-swap model has to blend a new face onto the original head, and that boundary is hard to get right. Check these areas before trusting a photo:
- Jawline and hairline. The single most reliable sign — look for a soft "seam" where skin texture or sharpness changes, hair that melts into the forehead, or a jaw edge that doesn't quite match the neck.
- Lighting direction. The face should be lit from the same direction as the rest of the scene. A face that is evenly lit while the scene has hard shadows is a red flag.
- Skin texture mismatch. Swapped faces often look smoother or more "airbrushed" than the ears, neck and hands around them.
- Accessories and edges. Glasses, earrings and hair strands crossing the face boundary tend to warp or break.
- Reverse image search. If the face was stolen from a real person, a reverse search may surface the original photo.
Automated analysis adds what the eye can't see: our AI image detector reviews signed Content Credentials, pixel-level signals and visual consistency in the original file, and shows each layer of evidence separately.
How to detect a face swap in a video
Video gives the fake more chances to slip up, because the swap has to hold up frame after frame:
- Blink patterns. Real humans blink every 4–6 seconds with varying duration. Swapped faces often blink at suspiciously regular intervals or in perfect two-eye synchrony.
- Profile moments. Most face-swap models are trained on frontal faces. When the head turns past roughly 45 degrees, watch for edge blurring, texture smearing or warping around the jaw and ear.
- Boundary flicker. Pause on fast motion — the blended region can lag or shimmer for a frame or two.
- Lip-sync drift. If audio was replaced too, mouth shapes may not quite match plosive sounds like "b" and "p".
Our deepfake video detector samples frames across the clip's full duration, reviews each detected face for swap artifacts, and cross-checks the result with an overall consistency analysis. Results show what was found and what remains uncertain. See also our guide on how to detect a deepfake video.
Face swaps in video calls and interviews
Real-time face swap has moved from research demos to a practical scam tool: virtual-camera apps can pipe a swapped face directly into Zoom, Microsoft Teams or Google Meet, and deepfake job candidates are now ranked among the top fraud threats for hiring teams. If you're on a call and something feels off, these live checks still break most current tools:
- The hand test. Ask the person to pass a hand slowly in front of their face. Real-time swaps struggle when a foreground object crosses the face plane — watch for the face flickering or the hand "cutting through" it.
- The profile turn. Ask them to turn their head fully to the side. Frontal-trained models smear or warp at the jaw, ear and hairline.
- Unscripted movement. Standing up, adjusting glasses, or holding an object next to the face all force the model outside its comfort zone.
These tests are a snapshot of mid-2026 — they will weaken as tools improve. For anything high-stakes, record the call and upload a clip to the video detector afterwards; boundary artifacts usually survive in the recording.
Who uses the face swap detector
- Recruiters and HR teams verifying that the person in a remote interview matches the candidate — before granting system access.
- Online dating users checking whether a match's photos or video calls use a borrowed face.
- Journalists and fact-checkers verifying viral celebrity or political clips before amplifying them.
- Fraud and KYC teams triaging suspected face-swap attacks on identity verification.
- Compliance teams preparing for the EU AI Act's transparency obligations for AI-generated content (Article 50, applicable from August 2026).
- Individuals documenting non-consensual face swaps of themselves for takedown reports.
Face swap detection FAQ
How accurate is the face swap detector? There is no honest single accuracy number across face-swap tools, edits and platforms. Original files preserve the most evidence; heavily compressed or re-encoded uploads are harder. Treat the result as strong evidence to weigh alongside context, not absolute proof.
Can it detect face swaps in both photos and videos? Yes. Photos go through the image pipeline (provenance, pixel-level signals and visual reasoning); videos are sampled across their full duration and each detected face is reviewed for swap artifacts, alongside a cross-modal reasoning pass.
Is this video real or AI? Can I check a video call recording? Yes. If you suspect a live video call used a real-time face swap, record a short clip and upload it to the video detector. Boundary artifacts around the jawline and hairline often survive in recordings.
Is the face swap check free? Yes — no signup is required. Service usage limits apply to keep the tool available for everyone.
What happens to my upload? Uploads are used only to analyze the current request and are not retained by DeepFakeCheck. We never train models on user uploads.
What can I do if my own face was swapped into an image or video? Keep the original file as evidence and run it through the detector to document the manipulation signals. In the US, the TAKE IT DOWN Act now obliges platforms to remove non-consensual intimate imagery, including AI face swaps, after a valid report.
Is the viral 'three-finger test' a reliable way to catch a live face swap? It still helps today: holding fingers or a hand in front of the face breaks many real-time face-swap tools, which is why the test went viral. But detection specialists already warn it is losing reliability as tools learn to handle occlusion. Use gesture tests as a first screen, then record a clip of the call and upload it to the video detector for an evidence-based check.
Does the EU AI Act require deepfakes to be labeled? Partly, from August 2, 2026: whoever deploys a deepfake must disclose that the content was artificially generated or manipulated (Article 50(4)). The separate duty for AI providers to embed machine-readable markings (Article 50(2)) applies from day one to tools launched on or after August 2, 2026, while tools already on the market before that date have until December 2, 2026. Scammers ignore labeling duties either way, so independent detection remains necessary. General information, not legal advice.