AI Image Detector is a multi-signal forensic screening tool that analyzes images using four independent detection models to determine if they are AI-generated.
What is AI Image Detector?
AI Image Detector is a web-based service that takes image files (JPEG, PNG, WebP up to 12MB) as input and produces a graded conclusion—from "AI-generated" to "No AI detected"—along with a detailed report showing each model's vote and evidence. It runs four independent detection engines in parallel, cross-checked by a published weighted-vote rule, and is available at aiimagedetector.info.
Key Features
- Four independent detection models — Sightengine GenAI, Winston AI, Was It AI, and SAFE (open-source, KDD 2025) each vote with published weights; all votes are named in the report.
- Batch screening — Upload up to 50 images at once; get a triage summary and CSV export.
- C2PA credentials and metadata trump card — Checks for embedded provenance credentials that override model votes when present.
- Graded conclusions — Five grades from AI-generated to real; disagreement among engines is reported honestly, never hidden behind a single percentage.
- Forensic chain-of-custody — Each upload records its SHA-256 hash; shareable report links and forensic PDF download available.
- Automatic credit refund — If a detection engine fails, its credit is automatically refunded; reports still complete with remaining engines.
- Free first report — Sign-up gives 4 welcome credits (one full report) with no credit card required.
Who is it for?
- Dating platform moderators — Verify profile photos to detect AI-generated faces before approving.
- Educators — Use evidence-graded reports to discuss potential AI-generated student submissions without accusations.
- Artist community moderators — Batch-screen uploaded artworks (up to 50 at once) to keep human-made galleries human.
- Journalists and researchers — Document AI detection passes with SHA-256 chain-of-custody for provenance verification before publication.
What can you do with it?
- Verify profile photos — Upload a dating or social media profile picture; get a shareable report link for moderators with each engine's vote.
- Screen digital art and illustrations — Run an image through all four models to see if it matches known generator fingerprints (Midjourney, DALL·E, Stable Diffusion, Flux, etc.).
- Check product and listing photos — Detect generated product shots before paying for them; batch upload up to 50 for marketplace safety.
- Verify news and viral images — Before sharing, run the image through the detector and get a graded conclusion with evidence laid out, not a percentage.
How does it work?
- Upload the untouched original image (or up to 50 in a batch). The service records its exact bytes and SHA-256.
- All four detection models run in parallel. If one engine fails, its credit refunds automatically and the report continues.
- A graded conclusion is produced—five grades from AI-generated to real—showing each engine's vote and the weighted consensus, with disagreement reported honestly.
Pricing
- Free — 4 credits on sign-up (first full report free).
- Pro — $19/month for 300 credits (~75 reports).
- Team — $59/month for 1500 credits (~375 reports).
A report costs 4 credits; per-report cost may increase if more engines are added (announced in advance).
FAQ
Can any AI image detector be 100% accurate?
No. This AI image detector cross-checks multiple detection models and reports disagreement instead of hiding it. It is designed to provide graded, evidence-based conclusions.
Why doesn't this AI image detector show a single percentage?
Because a single percentage hides which engines agreed or disagreed. Instead, every engine's vote is named in the report, and the weighted consensus is presented alongside the evidence.
What does 'Inconclusive' mean?
It means the engines disagreed significantly—some s
