Is this profile picture AI-generated?

Why made-up faces are everywhere

A generated face is free, unique and belongs to no one. Because the person in it doesn't exist, nobody can search for them, find them or complain that their photo was taken. That makes a made-up face the default disguise for fake accounts:

• networks of accounts that follow, like and share on cue;
• fake recruiters and "hiring managers" fishing for personal details, documents or payments;
• applicants who don't exist, or who aren't who they say they are;
• sellers and "experts" borrowing credibility from a friendly, professional face;
• throwaway accounts that pile into comment sections and reviews.

Fake accounts used to rely on stolen photos of real people. Many still do, and the two need different checks: a stolen photo has a history you can find, while a generated face usually has none. The checks below cover both, whether you are deciding to accept a connection, screening an applicant or reviewing a report as a moderator.

Step 1: look at the face, and around it

Early face generators left famous fingerprints, and most of them are gone. Today's images get skin, eyes and teeth right most of the time, so look at the places where the face has to meet the rest of the picture.

Treat any single oddity as a reason to keep checking, not as proof. Plenty of real people use a strange, blurry or heavily filtered photo as their avatar.

Step 2: check the account around the photo

A profile picture is one piece of evidence. The account wrapped around it is often more telling than the face itself:

• Age and activity: an account created recently, followed by a burst of connection requests or follows, deserves more care than one with years of ordinary posts.
• Other photos: real people turn up in different places, light, outfits and company over the years. A profile with a single perfect headshot, or a set of photos that all look like one studio session, is weaker evidence of a real person.
• The same person throughout: compare ears, hairline, moles, teeth and glasses across every picture. Generated faces drift slightly from image to image, even when someone asks for the same person each time.
• Mutual connections: who actually knows this person? A colleague you can ask is worth more than any photo.
• The story: a job at a real company that the company's own pages never mention, a city that doesn't fit the language or the hours they keep, a career that changes with each message.

Step 3: find out if the photo is stolen or made up

Fake profiles use one of two kinds of picture, and a different check catches each one:

1. 1

   Run a reverse image search

   Search the photo with Google Lens, Bing Visual Search or TinEye. If it turns up under another name, on a stock photo site or on an influencer's account, the photo is stolen, and you have your answer. Our reverse image search guide shows how.

2. 2

   Read a blank result carefully

   A generated face usually has no history, because it was made for this account. No results is what you would expect from a made-up face, and also from a private person who rarely posts photos. It is not proof either way.

3. 3

   Check the pixels

   This is where a detector helps most: it looks at the picture itself rather than its history. Drop the profile photo into Perspicax and read the answer alongside the rest of your checks.

4. 4

   Ask for something new

   A real person can send a photo taken today, holding something you name, or join a short video call. Someone hiding behind a made-up face usually finds a reason not to.

Keep the two outcomes apart. A stolen photo means the account is lying about who it is, and the person in the picture is a victim too. A generated face means the person may not exist at all.

How Perspicax checks a profile picture

Perspicax combines proven computer vision techniques with TrueEye 2.3, our state-of-the-art detector. It examines every region of the image, down to the pixels, and answers in one of three bands:

In testing, it caught 98.6% of images from the newest OpenAI, Google and ByteDance models on average, and wrongly flagged fewer than 1 in 2,000 original camera photos.

Give it the largest copy you can get. Avatars are usually shown small and squashed into a circle, so open the profile picture itself first: on most sites, tapping or clicking it shows it bigger. Save that version, or copy the image's address and paste it into Perspicax. The more of the picture there is, the more there is to check.

For recruiters and HR teams

Fake applicants and fake recruiters both lean on made-up faces, and a face is a weak identity check either way. Build the real checks into the process instead:

• Hold at least one live video interview, camera on, before any offer, and don't let it be the only check.
• Verify identity documents through your usual process, and compare the person on the call with the profile photo and the documents.
• Confirm employment history with the employers themselves, not only through contacts the applicant supplies.
• When a recruiter approaches your staff, check that they appear on their company's own careers or team pages, and that their email address uses the company's domain.
• Screen uploaded headshots for full AI generation at scale through the Perspicax API, and send flagged ones to a person for review rather than rejecting them automatically.

A flag from a detector is not, on its own, a reason to reject a candidate. Some people use an illustrated or AI-polished avatar for privacy. Ask about it, and judge the person on the checks that matter.

What to do if you think it's fake

• Don't share personal details, documents, codes or money with the account.
• Don't accuse anyone in public. A fake account may be part of a network, and the person in a stolen photo has done nothing wrong.
• Report the profile through the platform's own reporting tools, choosing fake account or impersonation.
• If the photo belongs to a real person, consider letting them know: a reverse image search often shows who they are.
• Keep screenshots of the profile and your messages in case you need them later.

Platforms are also starting to label accounts built around AI-generated people. Instagram, for example, has an "AI-generated profile" label for creators whose posts show people made with AI. A label like that is a disclosure, not a warning sign: the accounts to worry about are the ones that hide it.