Deepfake detector

Inspect the face and the image around it.

A realistic face can still sit inside an inconsistent file. AI Screener combines face-region analysis with pixel, replay, provenance, and—when a document portrait is present—identity consistency checks.

01

Face role detection

Faces are separated into presenter and document-portrait roles before relevant comparisons run.

02

Passive presentation clues

A research model checks a single frame for print, screen, or replay-like presentation signals.

03

Face-to-document comparison

When both roles exist, independent face embeddings compare the presenter with the ID portrait.

04

Localized manipulation

Noise, seam, recompression, and learned localization maps look for inconsistent regions around the face.

No automated detector can prove authenticity. AI Screener combines independent signals, shows the evidence, and leaves the decision with a human.

Run the free checker

Questions to ask of the result

Limits matter as much as the signals.

Does a face comparison prove identity?

No. Face similarity is research evidence for a human review and may be inconclusive when the portrait or selfie is low quality.