AI can now detect forged documents by checking both the images and the text — and explaining why it thinks so

Researchers built a system that catches AI-generated document forgeries by analyzing visual and textual evidence together instead of separately. This means document fraud detection moves from pattern-matching on images alone to something closer to how a forensic examiner would actually reason through a suspected fake.

Until now, document forgery detection treated the visual anomalies and the text as separate problems — you'd check if pixels looked wrong, then separately check if words made sense. This system checks them together in real time, catching subtle manipulations that slip through when you only look at one channel. The second piece matters more: the system has to show its reasoning in text, not just produce a confidence score. That shifts document forensics from 'the algorithm says fake' to 'here's why it's fake,' which is what courts and fraud investigators actually need.