Skip to content
perspicaxby avancer.ai (opens in a new tab)
How it works

How Perspicax reads an image.

Labels first, then every pixel. Four steps, a few seconds.

In short

How does AI image detection work?

Perspicax checks an image in four steps: it reads any Content Credentials (C2PA) or IPTC label, checks that the image is a photo, has TrueEye 2.1 examine the whole image region by region down to the pixels, and answers in three bands: likely AI-generated, inconclusive or likely real.

  1. Step 1 of 4: Provenance

    Read the label

    We look for Content Credentials (C2PA) and the IPTC digital source type. A signed label is the best evidence there is, and we show it before anything else.

    Provenance checks
    • Content Credentials (C2PA)Live
    • SynthID watermark checkSupported
  2. Step 2 of 4: Scope

    Check it is a photo

    Perspicax answers for photo-like still images. Screenshots, documents and graphics come back as outside scope, not as a guess.

  3. Step 3 of 4: Pixels

    Examine every region

    TrueEye 2.1 examines the whole image, region by region, down to the pixels, and looks for the traces image generators leave.

    Read the TrueEye 2.1 model card
  4. Step 4 of 4: Verdict

    Answer in three bands

    Likely AI-generated, inconclusive or likely real. When the evidence is not strong enough either way, we say so, and why.

Before you believe it, check it.

Your first scan is free, no account needed. A free account gives you 2 scans every day.

Scan an image