Start with the original photo
Screenshots, filters and repeated compression change the file the detector sees. Keep the earliest available copy and note where you received it.
Loading…
Upload JPG, PNG or WebP files up to 50 MB, or GIF files up to 20 MB. GIF checks analyze only the first frame. Get the detector result, available file information and suggested follow-up checks in one report, with a full PDF to download.
Screenshots, filters and repeated compression change the file the detector sees. Keep the earliest available copy and note where you received it.
An image can be generated, partly modified or simply resized. A detector signal does not identify every retouch, prove who made an edit or establish when the photograph was taken.
Compare related photographs, the original listing or publication, and available provenance. A low score cannot establish that the caption or the event is true.
Select up to 50 photos or a ZIP in your company workspace. Review each result and download individual reports with a CSV summary. No API integration is needed.
Batch image checksAdd image checks to your application. Receive structured results and track processing through job status or signed webhooks.
AI image & video detection APIShows whether the detector found image patterns associated with AI generation. The score is not a probability that the photo is fake.
Includes available format, dimensions and metadata. Missing camera details alone are not evidence of AI.
Explains what was checked, what remains unknown and how to investigate the image further.
Treat the result as evidence about detectable image patterns, not as proof of authorship. A model can be wrong in either direction, especially when an image has been resized, filtered or recompressed.
Important limitation: Screenshots, illustrations, 3D renders, very small images and heavily edited real photos may resemble generated content.
Read the full methodologyDemonstration only: this is not a customer case, live result or performance claim.
A striking JPG is downloaded from a social post, but the publisher cannot provide the original camera file.
An image-model indicator can be recorded alongside missing provenance and an unverified caption instead of being treated as the whole answer.
The detector output describes patterns in this copy. It does not prove who made the image or whether the depicted event happened.
Find the earliest copy, reverse-search the image and central subject, then compare the claim with independent sources.
Check an unusual image before sharing it.
Review product or profile imagery that looks synthetic.
Triage images that need source verification.
Upload the clearest original image you can find, review the detector signal, then compare it with reverse-image search, source history, metadata and any available Content Credentials.
It can identify patterns associated with AI generation, but it cannot certify that a photo is authentic or that the surrounding caption is true. Treat the result as one piece of evidence.
It can analyze a supported screenshot, but resizing, interface overlays and recompression can hide or introduce model patterns. The earliest original image is more useful whenever you can obtain it.
Upload JPG, PNG or WebP files up to 50 MB, or GIF files up to 20 MB. Only the first GIF frame is analyzed; the rest of the animation is not checked. Choose the clearest original copy available; screenshots, filters and repeated compression can affect the result.
A standard image check uses 1 credit. Optional source research adds 3 credits. Eligible personal accounts receive 5 welcome credits, and you can download the report as a PDF.
No. A standard check covers the detector result and available file information. You can add public-source research or run your own reverse-image search to investigate where the image appeared.
Use source history, provenance and independent checks before making a high-stakes decision.
Read the AI-image verification guide