isGenAI.com · Free deepfake check
Deepfake check: verify a suspected deepfake step by step
Start with what the file or page records: upload the image or video frame, or paste the post link, and isGenAI quotes any Content Credentials, AI label or saved prompt it finds. Then follow the steps below to trace where the media came from.
Free. No account. No face-swap score: only evidence you can retrace.
The free isGenAI detector
Check for AI. See the evidence.
Check my image for AI
Drop an image here or pick one from your device. The check starts the moment you choose a file.
PNG, JPEG, WebP, AVIF or single-page TIFF · Up to 20 MiBTIFF: full file evidence and eight pixel measurements, read straight from the original bytes, up to 40 million pixels.Every result shows exactly what we found and exactly how we checked it.
How to check a suspected deepfake
- Check the file’s own records. Run the original download, not a screenshot, through the checker above or the C2PA checker. A signed manifest or an IPTC “trainedAlgorithmicMedia” label that names an AI tool settles the question for that file.
- Pull stills from a video. Use the video frame extractor to save sharp frames, and the video metadata viewer to read the container’s creation tool and dates.
- Find the earliest upload. Search the frames with a reverse video search. An earlier, longer or higher-quality version often shows the original context or a creator who labelled it as AI.
- Look for a published fact-check. Viral deepfakes are often already investigated. isGenAI’s AI media cases link each verdict to the fact-checkers who made it.
- Check the claimed source. Look at the official accounts of the person or organisation shown. A denial or the real footage is stronger evidence than any visual impression.
What a missing record means
Most social platforms strip metadata on upload. In isGenAI’s own test, X removed every provenance record we uploaded (see the study). So “no AI record found” on a reposted clip does not mean the clip is real; it means the evidence has to come from source tracing.
Related
The AI video checker covers video links and files, the AI image detector explains each record type, and the Article 50 guide covers deepfake disclosure duties. The method and its limits are in the checker methodology.
Questions about checking deepfakes
Can isGenAI tell me if a video is a deepfake?
Not from how the faces look. isGenAI has no face-swap or lip-sync classifier. It reads the evidence a file or page records, such as C2PA Content Credentials, AI-generated labels and saved prompts, and the steps on this page help you trace the source. When a record says AI made or edited the media, the report quotes it.
Why not just use a deepfake detector score?
A score is an estimate that is hard to check and can be wrong on compressed, cropped or re-recorded copies. Fact-checkers usually combine detector scores with source tracing, earliest-upload searches and statements from the people shown. A recorded provenance fact or a traced original is something a reader can verify.
What counts as a deepfake under the EU AI Act?
Article 3(60) defines a deep fake as AI-generated or manipulated image, audio or video content that resembles existing persons, objects, places, entities or events and would falsely appear to a person to be authentic or truthful. Article 50(4) requires deployers to disclose it, with exceptions; see isGenAI’s Article 50 guide.
Is the check private?
Yes. Uploaded files are private by default, and nothing is published unless you ask for a public report.