The quickest way to spot an AI-generated profile picture is to open it at full size and check three places: the background, the ears and hair edges, and anything the person is wearing or holding. Then look at the account around the photo, because a fake face almost always sits on a thin, recent profile. A detector can back up your suspicion, but small, compressed avatars are among the hardest images to judge, so expect “Uncertain” more often than with a full-size photo.
Why fake accounts use AI faces #
A stolen photo of a real person can be caught by reverse image search. An AI face can’t, because it has never appeared anywhere before. That’s why fake recruiters, investment “advisers”, bot networks and romance scammers moved to generated faces. The FBI lists fake social media profile photos first among the ways criminals misuse AI images (FBI IC3).
People are also bad at catching them. In a 2022 study, participants identified AI faces only 48.2% of the time and rated them slightly more trustworthy than real ones, 4.82 versus 4.48 on a 7-point scale (PNAS).
Two kinds of AI faces, two sets of clues #
Older GAN faces #
Sites like This Person Does Not Exist made faces with a type of model called a GAN, and those faces still sit on a lot of fake accounts. Their clues:
- Eyes in the same spot. GAN models were trained on aligned portraits, so the eyes land at almost exactly the same height and position in every image. Line up a few suspicious avatars and the eyes overlap.
- Mushy backgrounds. Swirls of color that suggest a room or a park but contain nothing you can name.
- Mismatched accessories. One earring, two different earrings, glasses whose frames differ on each side.
- Hair that melts into the background or into a faint second outline.
- Odd teeth. One too-wide front tooth, or teeth that blur together.
Newer diffusion portraits #
Current generators produce faces that look like professional headshots. They usually avoid the GAN errors but have their own look:
- Studio lighting on what’s supposed to be a casual selfie.
- Skin with no pores, near-perfect symmetry and a faint glossy sheen.
- Backgrounds blurred so smoothly that no real object survives.
- Clothing with seams, buttons or logos that don’t quite resolve.
- A generic stock-photo quality: the person could be anyone.
AI headshot apps, which turn your selfies into a polished portrait, add a third category: a real person, rendered by AI. Those can trip a detector even though the account is genuine.
How to check the account, not just the face #
- Count the photos. Real people have messy histories: group shots, bad lighting, pets, old haircuts. One or two flawless portraits and nothing else is a warning sign.
- Check account age and activity. Joined recently, hundreds of connections, no posts of their own.
- Look for consistency. Is it clearly the same person across photos, with the same ears, moles and hairline? Generated sets often drift.
- Verify the job. Search the employer’s own website or staff page for the name.
- Reverse search anyway. It won’t find a fresh AI face, but it will find a stolen real one, which is a different red flag.
How to run an AI check on a profile photo #
Profile pictures are small. Platforms resize avatars heavily, and that throws away the fine texture a detector reads. So get the biggest version you can: tap the avatar to open it full screen, or open the profile in a browser and save the image.
On a phone, Expose AI does this in a few taps. Screenshot the enlarged photo, share it to the app, and use Crop to cut away everything around the face. You’ll get Likely real, Uncertain or Likely AI-generated with a confidence meter. The check runs on your phone, so the stranger’s photo isn’t uploaded anywhere. If the profile has several photos, Pro’s batch check takes up to 12 images at once.
“Uncertain” is an honest answer for a tiny image. The Pro model returns roughly half as many Uncertain verdicts as the standard one on the developer’s evaluation set, but some avatars simply don’t contain enough detail. When that happens, the account checks above matter more than any score.
What to do when you find one #
- Don’t act on requests. Ignore money, crypto, gift card or “job opportunity” requests from accounts you can’t verify.
- Report the account using the platform’s fake-account or impersonation option.
- Warn mutual connections if the fake is targeting your workplace or community.
If the profile belongs to a dating match, the stakes are higher; see how to check a dating match for AI photos. Marketplace sellers with AI faces often post AI item photos too, which we cover in spotting AI photos in marketplace listings.
Frequently asked questions #
Can LinkedIn profile pictures be AI-generated? #
Yes. Fake recruiter and sales accounts have used generated headshots for years, and AI headshot apps mean some real people now use AI-rendered photos too. Judge the account as a whole: work history you can verify, mutual connections you know, and posts written in a consistent voice.
Why can’t reverse image search find an AI face? #
Because the face was generated for that account and has never been posted anywhere else. Reverse search finds copies of existing images. It’s still worth running, since it catches accounts using stolen photos of real people.
Is using an AI profile picture illegal? #
Generally not. Many platforms allow AI avatars, though some ban impersonation or require an authentic identity. The problem is deception: pairing a fake face with a fake identity to scam people is fraud no matter how the photo was made.
Why does a detector struggle with small avatars? #
Detectors read fine patterns in texture and noise. An avatar that has been shrunk and compressed several times has lost most of them. Use the largest copy available, and treat “Uncertain” as a prompt to look harder at the account.