AI-content detectors are uncertain: how to verify a text or image

A detector score is not proof of authorship. Check provenance, original sources and context before making a claim or decision.

Adapted from a guide first written for Quebec readers: passages specific to Quebec were removed. Rules in your province may differ.

Automated detectors estimate whether material resembles patterns associated with generated content; they cannot reliably prove who made a particular text, image or audio clip. False positives and false negatives are possible, especially after editing, translation, compression or short samples. Do not use a detector score alone to accuse a student, employee or source of deception.

Verify the claim instead

• Find the earliest available copy and check the original publisher or author.
• Look for a source file, publication history, reliable context and independent corroboration.
• For an image, use reverse-image search and inspect whether the scene appears elsewhere with a different caption.
• For audio or video, contact the person through a separate trusted channel before acting on urgent requests.
• Record what you checked and communicate uncertainty clearly.

Metadata and AI-detection tools can provide clues, but metadata may be missing or altered. For school or workplace concerns, follow the institution’s review process and give the person a chance to explain. Report harmful or fraudulent content to the platform and, where appropriate, to police or a relevant regulator.

Canadian Centre for Cyber Security: identify misleading information · Version française →

Sources and limits

Source: Canadian Centre for Cyber Security. Detection tools are fallible and should not be treated as conclusive evidence of authorship.

General information based on federal and national rules; provincial rules may differ. Confirm with the official body before acting.