Detectors estimate. They never prove. We tell you exactly what a score can and can’t mean.
AbleProAI
Beyond text

Detecting AI Images, Audio, and Video

How to approach suspected AI images, voice clones, and deepfake videos: what detection tools can do, what visual tells no longer work, and practical checks that do.

The short answer

Visual “tells” like strange hands are no longer reliable. The best checks are provenance (where did it come from?), content credentials when present, and confirming through a separate channel, especially for urgent requests by voice or video.

Why spotting fakes by eye has stopped working

Early AI images had obvious flaws: extra fingers, melted text, mismatched earrings. Current image and video models have largely fixed those. Relying on visual tells now creates false confidence in both directions.

Images

  • Look for content credentials. Some cameras, editing apps, and AI tools attach signed records you can inspect with a content credentials viewer.
  • Run a reverse image search to find the earliest copy and its original context.
  • Check the source: who posted it first, and do they have a track record?
  • Use detection tools as one signal only. Like text detectors, image detectors produce probabilities and can be fooled by cropping, compression, or editing.

Audio and voice cloning

  • A few seconds of recorded speech can be enough to clone a voice convincingly.
  • Scams often impersonate a family member, boss, or bank and create urgency: money now, secrecy, unusual payment methods.
  • Hang up and call back on a number you already know.
  • Agree on a family code word for emergencies.
  • Businesses should require a second approval channel for payment or account changes requested by phone or video.

Video

  • Check whether the event is reported by reliable outlets.
  • Look for the original upload rather than reposts.
  • Be cautious with clips that are short, low resolution, or emotionally charged, which are the hardest to verify.
  • For live video calls, ask the person to do something unscripted, or confirm through a separate channel.

The bigger picture

As generated media gets better, the answer moves from detecting fakes to proving what is real: signed content credentials from cameras and apps, verified sources, and records of how something was made. The same shift is happening with writing.

Last reviewed October 2026 by the AblePro team. Found something out of date? Tell us.