AI Detector for Journalists
Updated
An AI detector for journalists is most useful when a viral image or clip lands on the desk. Upload a PNG, JPG or WEBP image up to 15 MB for an AI-generation score plus a separate face-manipulation score, or an MP4, MOV or WEBM clip up to 50 MB for a video score with ML, OCR and metadata sub-signals.
Limitations: Every result is a probability that can be wrong, not proof of how an image was made. How results can be wrong →
3 free scans in total, shared across all five tools (not per day or per month). No signup needed to start.
How an AI Detector for Journalists Fits Verification
- Start with provenance: who posted it first, when, and where; a reverse image search often answers more than a detector.
- Run the detector: use the score as one signal alongside location, weather, shadows and other visual checks.
- Get the original file: screenshots and re-uploads lose detail and metadata, which weakens any automated result.
- Text claims: statements and press releases of 300+ words can be run through the text detector.
What the Scores Mean
The image check returns two separate probabilities: whether the image looks AI-generated, and whether a face looks manipulated (a deepfake signal). The video check returns an AI-generation probability with ML, OCR and metadata sub-signals; it is not a dedicated face-swap detector.
Neither score is proof. A low score doesn't clear an image, and a high one doesn't settle a story. Treat them like any other single source: corroborate.
Reporting Responsibly
If you report that content is likely AI-generated, explain how you reached that view, and avoid presenting a tool's percentage as a verdict. Readers and the people pictured deserve the uncertainty to be clear.
Only upload material you have the right to handle; uploaded files are sent to the third-party providers listed on the How It Works page.
A Verification Checklist for Images
Before you rely on any score, note where the image first appeared, who shared it, and whether the same scene appears from other angles. Check visible details against known facts: weather, signage, uniforms, time of day.
Then run the image check on the best copy you can get, record both scores, and keep a note of what each other check showed. That record makes your conclusion easier to explain to editors and readers.
Frequently asked questions
If you have the file (MP4, MOV or WEBM, up to 50 MB). Downloaded or screen-recorded copies lose detail, which makes results less reliable.
It returns a separate face-manipulation score alongside the AI-generation score. Both are probabilities, not proof.
The image check looks at whether the image itself is AI-generated, not whether the text in it is true.
Use it as one input. If you mention it, explain its limits rather than presenting it as a verdict.