Step 01
Read the pixel signal
The model looks for visual patterns associated with fully generated and AI-edited imagery. It does not rely on filenames or EXIF to produce the score.
Evidence before certainty
AI photo detectorfor a second look.
Check whether a photo may be AI-generated or AI-edited. Get a clear model score, useful metadata clues, and the limitations that other checkers often hide.
Deleted after analysis Never presented as proof
Visual origin review
JPG · PNG · WEBP
Checking today's free allowance…One moment
Resets 00:00 UTCDrop in a photo for a second look
Drag and drop, paste an image, or choose a file. Your original is temporarily transferred and deleted after analysis.
Maximum 10 MB · No sign-up for free checks
No long-term image retention
No training use by us
Probability, not proof
Signal 01
Pixel model
A visual AI-generation signal from 0–100%
Signal 02
Metadata context
Camera and editing clues when the file contains them
Signal 03
Honest uncertainty
A middle band instead of a forced yes or no
How the verdict is built
One answer, with the evidence left visible.
Most AI image checkers give a percentage and stop. OriginVerdict shows what the number can say, what the file adds, and where both can fail.
Step 02
Surface safe file clues
Camera make, camera model, capture time, dimensions, and editing-software tags are shown when present. GPS data is never returned.
Step 03
Give a restrained verdict
Scores in the middle stay inconclusive. Every result explains failure modes so a detector never becomes the sole basis for a consequential decision.
Read the score correctly
A probability is a clue.
Not a confession.
AI detection is an adversarial, moving target. Re-encoding, screenshots, image edits, and newly released generators can move a score in either direction. The source, publication history, and corroborating evidence still matter.
0–20%
Low AI signal
Few visual patterns associated with generation. This does not authenticate the photographer or source.
21–79%
Inconclusive
The image sits in the uncertainty band. Avoid a binary conclusion and look for more context.
80–100%
High AI signal
Patterns resemble generated or AI-edited imagery. This remains a model assessment, not forensic proof.
Who it helps
For people who need to pause before they publish.
Journalists & researchers
Triage a visual before contacting the source or seeking stronger forensic evidence.
Marketplace reviewers
Flag suspicious product, profile, or claim imagery for a closer manual review.
Educators & students
Discuss synthetic media literacy with a result that shows uncertainty, not false certainty.
Trust & safety teams
Add a quick visual signal to a broader moderation workflow without making it the final arbiter.
Calibration gallery
Known samples, checked through the same pipeline.
These are saved test results from July 21, 2026, not a promise of universal accuracy. Provider updates, re-encoding, and edits can change a score. Sources and licenses are linked on every camera sample.
99% AI signal
99% AI signal
0.1% AI signal
0.1% AI signalThe file is examined, then deleted.
The original passes through private temporary storage and is deleted after analysis. GPS parsing is disabled, provider raw responses are not kept, and only a daily quota counter is stored. Sightengine’s processing terms also apply.
Third-party processing
disclosed in privacy policy
Simple credits
Start free. Pay for checks, not mystery features.
The paid prices below are the planned launch structure. Checkout remains disabled until the dedicated Stripe account is connected and tested.
Free review
$0forever
For a quick second look before you share, cite, or moderate an image.
- 2 image checks per day
- AI signal score
- Safe EXIF clues
- Clear limitations
Available now
Starter
$7.99200 credits
For a burst of source checks, marketplace review, or research.
- 200 image checks
- One-time purchase · valid 1 year
- Same evidence-rich result
- No automatic renewal
Billing setup in progress
Pro
$19.99500 credits
For larger review queues without committing to a subscription.
- 500 image checks
- One-time purchase · valid 1 year
- Same evidence-rich result
- No automatic renewal
Billing setup in progress
Questions worth asking
Before you trust a detector.
01What is an AI photo detector?
+
An AI photo detector estimates whether visual patterns in an image resemble content produced or edited by generative AI. OriginVerdict also surfaces safe file metadata so the score is not presented without context.
02Is an AI image detector 100% accurate?
+
No. No detector can prove origin from pixels alone. Compression, screenshots, retouching, unusual camera pipelines, and newly released generators can all affect a score. Treat it as one signal in a broader review.
03Can it detect Midjourney, DALL·E, Flux, and Stable Diffusion?
+
The underlying model is designed to recognize patterns from major current generators and is updated over time. A high score still does not prove which specific generator was used.
04Can it detect an AI-edited real photo?
+
It can surface a generative-AI signal in fully generated and AI-edited images, but subtle edits are harder than fully synthetic images. The result deliberately uses an uncertainty band instead of forcing every image into yes or no.
05Do you keep uploaded photos?
+
OriginVerdict temporarily transfers the original through private storage and attempts to delete it immediately after analysis. A one-day automatic cleanup rule covers interrupted requests. The file is also sent to Sightengine, so Sightengine's processing and retention terms apply. See our Privacy Policy before uploading sensitive material.
06Is “AI photo detector” different from “AI image detector” or “AI image checker”?
+
People use those phrases for the same basic intent: checking whether a visual may be AI-generated. We use one canonical tool page so the explanations, limitations, and result experience stay consistent.