For commerce
A customer sends damage photos? Verify their authenticity before the return leaves their hands.
ALTERTRACEIMAGE FORENSICS · SIX INDEPENDENT LAYERS
Evidence,
not instinct.
Find out whether a product photo was generated or locally altered with AI — before you approve a refund, reject a claim or question genuine evidence.
● Cases auto-delete after 7 days⌁ Shareable PDF report○ No subscription
For consumers
A shop questions your photos? Send a clear, independently generated forensic report.
Beyond claims
AlterTrace works wherever a photo can no longer be accepted as truth on sight.
START ANALYSIS / 01
One upload.
A defensible answer.
The production stack combines file-level forensics with independent commercial detection models. No single black box gets the final word.
from 19 PLNUp to 5 images now + 2 follow-ups within 7 days
Drop evidence here
JPEG, PNG or WebP · up to 5 now + 2 follow-ups · max 20 MB each
C2PA EXIF XMP ELA NOISE ENTROPY MICROTEXTURE SHA-256 HIVE SIGHTENGINE TRUTHSCAN REALITY DEFENDER HEATMAP SERIES CONSISTENCY
DETECTION STACK / 06
One verdict. Multiple independent signals.
The production stack combines file-level forensics with independent commercial detection models. No single black box gets the final word.
AlterTrace Agent
C2PA · EXIF · XMP · JPEG/PNG structure · generator signatures · SHA-256 · ELA · noise · entropy · microtexture · Case Coherence
Hive
Synthetic-image probability and likely generator family
Sightengine
AI generation, manipulation and deepfake signals
AI or Not
AI generation, deepfake and generator-family classification
TruthScan
Local manipulation analysis and provider heatmap
Reality Defender
Multi-model deepfake and AI manipulation consensus
CALIBRATED VERDICT
A score built to be
understood, not worshipped.
0–24 Minimal
25–49 Elevated
50–74 High
75–100 Very high
KNOW BEFORE YOU TRUST / FAQ
Questions deserve precise answers.
01Can the result prove fraud?+
No. AlterTrace estimates the likelihood and identifies technical signals. It supports a decision; it does not establish intent or replace an expert witness.
02Can it detect a small AI edit?+
That is a core use case. Local edits can leave inconsistent noise, compression, metadata or microtexture patterns, but subtle edits are harder than fully generated images and no detector can promise certainty.
03Why use six layers?+
Detectors fail differently. We combine provenance, file structure, pixel forensics and independent model scores to reduce reliance on any one vendor.
04What does the overall score mean?+
It is a calibrated risk indicator, not a percentage of pixels changed. The report separates minimal, elevated, high and very high likelihood and explains the strongest evidence.
05Are original files important?+
Yes. Screenshots and messenger recompression remove metadata and weaken forensic traces. Always upload the original file when possible.
06How long are files stored?+
The complete case is automatically and permanently deleted after 7 days. During that window, up to two follow-up images can be added to the original five.
07Will a clean result guarantee authenticity?+
No. A clean result means the available signals did not establish likely AI involvement. It must never be presented as an absolute guarantee.
08Can I analyse several photos from one claim?+
Yes. A 19 PLN case includes up to five initial images and two follow-ups within seven days. Case Coherence checks that every image shows the same product and claim before paid detection begins.
09Who builds the external detectors?+
Hive, Sightengine, TruthScan and Reality Defender are independent providers. Their outputs are preserved separately in the report and are not presented as AlterTrace's own models.
10Is the PDF suitable for sending to a customer?+
Yes. It uses neutral language, lists methods and limitations, and includes the file hash so the analysed evidence can be identified later.