AI face check
Fake profile picture checker
Scam accounts, catfish and fake recruiters increasingly use AI-generated faces instead of stolen photos, because a generated face can’t be found by a reverse image search. Drop the profile picture to check it.
Free · no sign-up · your file never leaves your device
Drop or paste the profile photo, click to choose, or paste
JPEG · PNG · WebP · AVIF · HEIC (Safari) — checked on your device
Why fake profiles switched to AI faces
For years the standard advice for spotting a romance scammer or fake recruiter was to reverse-search their photo: scammers stole pictures of real people, and the search would turn up a model’s Instagram or a stock-photo page. Generators made that advice obsolete. A face from StyleGAN, Stable Diffusion, FLUX or a dedicated “AI headshot” app is unique, so the search comes back empty — which ironically makes the profile look more trustworthy.
AI faces are now common in pig-butchering investment scams, fake job offers, sock-puppet accounts in political influence campaigns, and bot networks that boost products or posts. LinkedIn, X and dating apps remove large numbers of them, but plenty get through.
What GPTTrace finds in a profile picture
Profile photos are usually small and recompressed by the platform, so metadata is almost always gone. That leaves the pixels. The neural classifier in GPTTrace was trained on output from thousands of generators, including face-specialised models, and works on the texture statistics of skin, hair and background rather than on how the face looks. It usually still has signal at profile-picture sizes, but the smaller and blurrier the image, the more cautious its score.
Two practical tips improve the result:
- Get the largest version. Platforms serve a thumbnail by default; opening the photo on its own often gives a bigger file.
- Check every photo on the profile. One inconclusive image means little. Four photos that all score high, with no camera metadata and identical lighting, is a strong pattern.
Signs beyond the picture
A detector is one tool. Fake profiles also give themselves away through behaviour:
- They move the conversation quickly to WhatsApp, Telegram or another private app.
- They are always unable to video call — the camera is broken, the connection is bad, they are working on an oil rig.
- Their story is glamorous but vague: an investor, a surgeon abroad, a recruiter for an unnamed client.
- Money enters the conversation, often as an investment opportunity or an emergency.
If a match agrees to a live video call, that is still worth more than any image check — though real-time face-swapping exists, so watch for glitches around the edges of the face when they turn their head or pass a hand in front of it. Our deepfake detector page covers those signs.
How accurate is it? Our measured numbers
We test GPTTrace on labelled image samples and publish the results, including where it does badly. It is tuned to keep false accusations rare, so it misses some AI content rather than flag real work.
image check: AUC 0.938 (cross-validated)
679 labelled samples (399 AI, 280 human), run 2026-10-08. At the “Likely AI” line it caught 64% of AI samples and wrongly flagged 5% of human ones.
| Source | Truth | Samples | Result at “Likely AI” |
|---|---|---|---|
| lfw-faces | Human | 40 | 0% wrongly flagged |
| ffhq-faces | Human | 40 | 0% wrongly flagged |
| celeba-faces | Human | 40 | 18% wrongly flagged |
Data sources and method: methodology & accuracy.