VeriSight

AI-Powered Image Authenticity Analysis

See Beyond the Image.

VeriSight uses AI to analyze images, identify potential signs of manipulation or AI generation, and support evidence-based reviews.

Three checks

How it works
AI likelihood
A 0 to 100% score in one of four bands
Tamper check
Regions edited with image software, outlined
Visual review
Four views to inspect side by side
Illustration of a receipt photo being scanned, revealing the analysis view beneath it with one flagged region outlined.

Sample result

AI likelihood 92%
Very likely AI-generated or AI-edited
Tamper check
1 region flagged
Visual review
4 views ready

What you get

One report. Three independent checks.

Every analysis ends in a report your team can read, print and file. It records exactly which file was checked and sets out what each check found, without folding them into a single verdict.

Open the sample reportAnalyze an image
verisight.tech

Image Authenticity Report

VeriSight

Report ID

5f874c99

The analyzed image, sample-ai-generated-cityscape.jpg

File record

File name
sample-ai-generated-cityscape.jpg
Analyzed
Oct 10, 2026, 17:29 UTC
Dimensions
1200 × 600 px
SHA-256
249f28ec…32bec
Method version
VeriSight 0.1.0

Findings at a glance

01  AI generation

99%

Very likely AI-generated or AI-edited

02  Tamper check

0 regions

No edits detected

03  Visual review

4 views

Not scored

01

AI generation

The detector found strong signs of AI generation or AI editing in this image.

The sample report, on a public-domain, AI-generated picture (Dennis Sylvester Hurd, CC0), analyzed as published. The values are what the checks returned.

  1. A record of the file

    Report ID, time of analysis and the file's SHA-256 hash, so a printed report can be matched to the original.

  2. AI likelihood

    A score from 0 to 100%, placed in one of four bands from no AI signal to very likely.

  3. Tamper check

    Regions edited with image software, outlined on the image and listed with their confidence.

  4. Visual review

    Four views computed from the image, each with a note on how to read it.

How it works

Three steps, a few seconds

  1. Step 1

    Upload a photo

    JPEG, PNG or WebP, up to 4 MB. The image stays in memory for the length of the analysis and is never stored.

  2. Step 2

    Run three checks

    One model scores the image for AI generation or AI editing, and a second outlines regions edited with image software. VeriSight then computes four views of the image that make typed-over text and pasted regions easier to spot.

  3. Step 3

    Review the evidence

    An AI likelihood score placed in one of four bands, flagged regions drawn on the image with their confidence, and four views to inspect side by side with notes on how to read them. The report can be printed.

Reading a result

What the three checks mean

The AI check returns a likelihood between 0 and 100% that falls into one of four bands. The tamper check either finds no edited regions or outlines the ones it found, each with a confidence. The visual review gives you four views computed from the image, each with a note on how to read it. All three are starting points for a reviewer, not a decision.

AI likelihood

  • 0–14%
    No AI signal detected

    The detector found little or no evidence of AI generation or AI editing in this image.

  • 15–49%
    Weak or inconclusive AI signal

    The detector found a weak signal. It is not strong enough to conclude that the image is AI-generated or AI-edited.

  • 50–84%
    Likely AI-generated or AI-edited

    The detector found clear signs of AI generation or AI editing. Corroborate with other evidence before drawing conclusions.

  • 85–100%
    Very likely AI-generated or AI-edited

    The detector found strong signs of AI generation or AI editing in this image.

Tamper check

  • No edits detected

    The tamper model found no region that looks edited with image software. Careful edits can still go unnoticed; use the views to check fields that matter.

  • Edited regions detected

    The tamper model found one or more regions that appear to have been edited with image software, such as pasted or typed-over content. Each region is outlined on the image with its confidence.

  • Not run

    The tamper check produced no result for this image. The reason is shown in the report; the other checks are unaffected.

Visual review

  • Cyan ink channel

    CMYK cyan channel, brightness = ink amount (black = none, white = full). Fields printed in one pass share one tone; a field that is markedly brighter or darker than comparable fields was produced differently.

  • Magenta ink channel

    CMYK magenta channel, brightness = ink amount. Read it like the cyan channel and compare text fields of the same kind.

  • Ink color map

    Dark strokes with their color cast amplified five times; everything else faded. Text written in one pass shares one cast.

  • Error level analysis

    The image re-saved as JPEG at quality 95 and the change in each pixel amplified. Regions saved the same number of times at the same quality change by similar amounts, so a text field or patch that is markedly brighter or darker than comparable content, or that shows a different block pattern, carries a different saving history. Bold and high-contrast strokes are always brighter than light ones, so compare like with like.

Limits you should know

Where detection falls short

  • A score is a probability estimate, not proof. Treat it as one piece of evidence and corroborate before acting on it.
  • The tamper check looks for regions edited with image software, such as pasted or typed-over fields. It is a separate model from the AI check: it can miss careful edits and can flag heavy recompression or scanning artifacts.
  • The views show where to look, not whether the image was edited. Bold or high-contrast text is always brighter in the error level view, and personalized fields printed on a different printer can differ in the channel views without any edit. Compare like with like.
  • Images sent through messaging apps or email are recompressed and lose detail. When you can, analyze the original file.
  • Screenshots and heavily downscaled images carry fewer traces of generation or editing, so results are less reliable.
  • Face swaps and other manipulated faces are not covered by this analysis.
  • Detection keeps pace with new generators only through continuous model updates. A low score on output from a brand-new generator is possible.

Privacy

Nothing is kept

Each image is held in memory only for the duration of the analysis. During analysis it is submitted to the detection models and is not retained by them or by VeriSight afterwards. VeriSight has no database and no file storage; a report exists only in your browser unless you print it. Read the privacy policy

Analyze an image