Compression and screenshot-like artifacts reduce certainty.
Probability, confidence, metadata, C2PA, limitations
AI Image Detector
Check AI Images With Evidence
Upload a JPG, PNG, or WebP image and get AI probability, confidence, metadata clues, C2PA hints, forensic signals, and clear limitations. Built for people who need more than a blind yes/no score.
No signup required for anonymous checks · JPG, PNG, WebP · Privacy-first processing
Probability estimate, not proofNo public image galleryCloudflare-first infrastructure
Drop an image here, or choose a file
JPG, PNG, or WebP · 5MB recommended · private by default
We return a probability estimate, confidence, signals, and limitations.Sample result · Uncertain
Evidence-based image check complete
This result is a probability estimate, not proof. AI image detectors can produce false positives and false negatives.
AI Probability74%
Likely camera/originalUncertainLikely AI-generated
Texture repetition and lighting smoothness need independent review.
No camera data found. Missing metadata does not prove AI generation.
No credential signal was detected in this file.
Definition
An AI image detector is a tool that helps people check whether a photo or picture may have been generated or manipulated by artificial intelligence.
Good detector results should explain probability, confidence, metadata, C2PA credentials, forensic signals, and limitations instead of returning a blind yes/no answer.
Result details stay visible, including uncertainty.
The interface is built around explainable evidence cards, not a single magic verdict. Labels stay limited to likely, uncertain, or unlikely.
Sample result · Uncertain
Evidence-based image check complete
This result is a probability estimate, not proof. AI image detectors can produce false positives and false negatives.
AI Probability74%
Likely camera/originalUncertainLikely AI-generated
Compression and screenshot-like artifacts reduce certainty.
Texture repetition and lighting smoothness need independent review.
No camera data found. Missing metadata does not prove AI generation.
No credential signal was detected in this file.
Use source checks, reverse search, and human review for important decisions.
Combine probability with provenance, metadata, C2PA, and context.
Probability, not proof
Every result carries a visible false-positive and false-negative warning. No legal, academic, or identity determination is implied.
One signal in a review stack
Use source verification, reverse image search, metadata review, C2PA credentials, and human review for important decisions.
How it works
A simple detector flow with conservative copy and no overclaiming.
Upload privately
Choose a JPG, PNG, or WebP. The preview keeps privacy and limitations copy close to the file picker.
Read evidence cards
Review probability, confidence, primary signals, metadata, C2PA hints, and limitations.
Decide responsibly
Use the result as supporting evidence, then verify source, context, and provenance.
Scenario workflows, not generic feature cards.
Each use case keeps the same safety boundary: detector output is a supporting signal, never the only basis for action.
Ecommerce & Marketplace Review
Seller images look polished but the source is unclear.
Review probability, metadata, C2PA hints, and risk notes before moderation.
Do not reject a seller on detector output alone.
Newsroom & Fact-Checking
A submitted image is spreading quickly across social channels.
Use the score as an early triage signal before source and reverse-image checks.
Treat every result as supporting evidence, not a final call.
Profile Picture Risk Review
A profile image appears synthetic or edited.
Check for possible generation signals without claiming identity verification.
Never label a person as fake from one image score.
Deepfake-Risk Images
An image may have been generated, edited, or context-shifted.
Combine detector output with provenance, metadata, and human review.
Use “possible signals” language; avoid sensational claims.
Accuracy and privacy are part of the interface.
Limitations are not buried in legal pages. They appear in the result, accuracy section, privacy flow, and scenario pages.
False positives and false negatives happen
Screenshots, compression, resizing, edited images, and missing metadata can reduce confidence. Important reviews need more than one detector result.
Upload → transient processing → metadata result → short TTL/delete
No public gallery. Uploaded images are not positioned as training data for our own models.
Usage-cap pricing, with P0 waitlist language.
Pricing cards show limits first. Pro is highlighted but not presented as live Stripe checkout.
$0
3 checks/hour/IP
Check My Image — Free$0
20 checks/month
Join Free Account List$19/month
1,000 checks/month
Join Pro Waitlist — 1,000 Checks/Month$49/month
5,000 checks/month
Request API Access$99 once
12 months Pro benefits · max 50 seats
Join Lifetime Early ListP0 plan language is waitlist/request intent until batch reports, payment, legal docs, cost controls, and success rates are production-ready.
FAQ
Conservative answers for high-risk interpretation, privacy, and accuracy questions.
Can an AI image detector prove an image is fake?
No. The result is a probability estimate, not proof. Use it as one signal alongside source checks, reverse image search, metadata review, C2PA credentials, and human judgment.
What image formats are supported?
The P0 interface is designed for JPG, PNG, and WebP uploads, with 5MB recommended for fast checks.
Why can screenshots reduce confidence?
Screenshots, compression, resizing, and edits can remove metadata and introduce artifacts that make image-origin signals less reliable.
Do missing EXIF metadata or C2PA credentials prove AI generation?
No. Missing metadata is common and does not prove AI generation. It is only a supporting signal.
Are uploaded images public?
No public gallery is part of this product direction. The UI is designed around private, transient processing.
Can I use this for legal, academic, or identity decisions?
No. The page avoids legal proof, academic punishment, and identity verification claims. Important decisions require independent review.
What does “Uncertain” mean?
It means the image has mixed or limited signals. The safest next step is to collect more evidence before acting.
Will there be batch checks?
Batch checks and reports belong to the Pro/P1 direction and should stay as waitlist or request-access CTAs in P0.
Is the checker free?
Anonymous checks are framed around a small free usage cap. Account and Pro plans add higher monthly limits.
Does this replace a human reviewer?
No. It is a workflow aid that makes signals visible; it does not replace human review or source verification.
Browse all image authenticity guides
Move between the tool page, responsible-use guides, accuracy notes, and scenario pages without losing the probability-not-proof framing.