Why eyeballing images no longer works
Early AI images failed on anatomy and fine detail, which trained everyone to look there first. Current models render hands, faces, and text cleanly enough that those checks pass, so a confident 'that looks real' tells you very little now.
Some generators leave a trail you can read instead of squinting at pixels: tools like DALL·E stamp images with C2PA content credentials that say plainly that AI made them. verif·ai reads those credentials. A credentialed image gets an authoritative flag. An image whose metadata has been stripped, or that never carried any (most Midjourney and local Stable Diffusion output), comes back as 'unverified' rather than a guess.
Reading verif·ai's image result
verif·ai outlines images it judges likely AI-generated with a dashed border and attaches a confidence score out of 100 plus a high / medium / low level. The strongest signal comes from embedded content credentials. When an image still carries the C2PA metadata that generators like DALL·E stamp in, that's near-certain confirmation. When the credentials are missing, verif·ai marks the image 'unverified' rather than scoring it from appearance alone.
Because the outline sits on the image itself, you can scan a whole page of photos at once and see which ones are flagged without checking them one by one.
how it works
- 01
get verif·ai in your browser
Once verif·ai is in Chrome it's ready on any page you read.
- 02
open the page
Go to the page with the image you want to check.
- 03
hit scan
Open verif·ai and scan. It evaluates the images on the page on-device.
- 04
read the outline
Likely AI images get a dashed outline and a confidence score out of 100.
frequently asked
Can verif·ai check any image on a page?
It scans the images on the page you're viewing when you hit scan and outlines the ones it judges likely AI-generated.
What does the confidence score mean?
It's an estimate from 0 to 100 of how likely the image is AI-generated, paired with a high/medium/low level. Higher means stronger suspicion, not certainty.
Does it upload the image somewhere to check it?
No. verif·ai runs on-device in your browser and keeps no server-side history of what you scan.
Will it catch every AI image?
No detector is perfect, and verif·ai can't pretend otherwise. It gives a confidence read rather than a guarantee, so treat a low score as 'uncertain' rather than 'confirmed real'.
Last updated June 7, 2026