AI Image Detection Tool For Journalists

For newsrooms, the cost of getting an image wrong runs both ways: publishing an AI fake, or wrongly dismissing a real photo as fake. No single tool settles it, so verification teams combine several, with documented steps.

A practical toolkit

• Metadata and provenance: AI Image Detector (made by Patrick Bushe, who runs this site) reads EXIF, IPTC, XMP and PNG data in Chrome and flags AI signals without uploading the image. Content Credentials Verify checks signed Content Credentials, which some news agencies and camera makers now use for real photos.
• Reverse image search: Google Lens, TinEye, Bing and Yandex each index different images. Use several.
• InVID-WeVerify plugin: a free browser plugin built for journalists, bundling reverse search, metadata, magnifier and forensic filters.
• Visual detectors: services such as Hive return a probability; useful as one input.
• Geolocation: Google Earth, street-level imagery and maps to match landmarks.

A workflow

1. Get the original file and ask the source how and when it was taken.
2. Read metadata and Content Credentials.
3. Search for earlier copies.
4. Check the scene: weather, shadows, signage, landmarks.
5. Look for visual artifacts.
6. Run a detector if still unsure.
7. Have a second person review before publication.

Reporting uncertainty

If you can't confirm an image, say so: "could not be independently verified." If you publish analysis of a suspected fake, explain what you checked rather than relying on a detector score, which readers can't evaluate.

Know the limits

Detectors can mislabel real photos, especially compressed or heavily edited ones, and can miss images from newer generators. Metadata is usually stripped on social media. Provenance and source reporting remain the strongest evidence.

See AI image detection for newsrooms and deepfake images in news.