How To Detect AI Generated Memes

Memes are the hardest images to check for AI. They're screenshotted, cropped, captioned and re-shared so many times that any trace of how the original was made is usually long gone. Still, a few checks help, especially when a meme presents something as real.

When it matters

A joke image that's obviously absurd doesn't need checking. It matters when a meme claims something happened: a politician doing something, a disaster scene, a celebrity quote with a photo. Those are the ones worth a minute before you share.

Trace the origin

1. Crop out the caption and run the image through Google Lens or TinEye.
2. Sort by oldest results if the tool allows, to find the first appearance.
3. Check meme databases and fact-checking sites, which often track viral images.

AI memes often trace back to an AI art community, a known AI account or a satire page.

Visual clues that survive sharing

• Crowds where faces blur into each other.
• Hands, limbs or objects that don't make sense.
• Signs and text in the scene, not the caption, that are garbled.
• A polished, cinematic look on something supposedly snapped on a phone.

Don't expect metadata

A metadata checker such as AI Image Detector (made by Patrick Bushe, who runs this site) will almost always say "No Metadata Found" on a meme, because sharing strips the file. That result is inconclusive. If you find the original upload, check that file instead.

Detection services

Visual AI detectors, such as Hive's, give a probability score from the image's pixels. They can be useful on memes, but heavy compression, captions and cropping reduce their accuracy. Treat the score as a hint.

See checking viral images and checking images before sharing.