Why you can't just tell by looking
Current image models have mostly fixed the giveaways people learned to spot: extra fingers, warped text, a melting background. A picture that passes those checks isn't cleared. It just means the easy tells are gone and your eye has run out of things to catch.
The most reliable evidence sits outside the pixels, in the C2PA content credentials some generators embed. verif·ai reads them, so a DALL·E image that still carries its credentials is confirmed AI, while a screenshot, a re-upload, or a generator that adds none (Midjourney, locally-run Stable Diffusion) comes back 'unverified' instead of a confident-sounding guess.
What verif·ai tells you, and what it doesn't
Scan the page and verif·ai outlines the images it judges likely AI-generated with a dashed border, plus a confidence score out of 100 and a high/medium/low level. A high score is a strong reason to be sceptical of that image. A low or medium one means uncertain, not confirmed real.
It's a confidence signal, not a verdict. No detector is right every time, so verif·ai gives you a calibrated read rather than a guarantee. Use it to decide how much to trust an image, not as proof either way.
how it works
- 01
open the page
Stay on the page with the image you're unsure about.
- 02
hit scan
Open verif·ai and scan. It checks the images on the current tab on-device.
- 03
look for the outline
Images it judges likely AI-generated get a dashed border.
- 04
read the score
Each flagged image carries a confidence score out of 100 and a high/medium/low level.
frequently asked
How do I check a specific image on a page?
Open verif·ai on that page and hit scan. It evaluates the images on the current tab and outlines the ones it judges likely AI-generated, so you can see which one is flagged.
Does a high score mean the image is definitely AI?
No. A high score means strong suspicion, not certainty. verif·ai reports a confidence level rather than a verdict, so treat it as a signal to dig deeper.
Does the image get uploaded anywhere to check it?
No. verif·ai runs on-device in your browser when you scan and keeps no server-side history of what you check.
If the score is low, is the image real?
Not necessarily. A low score means uncertain rather than cleared. No detector catches every AI image, so a low result isn't proof the image is genuine.
Last updated June 8, 2026