How to spot a fake or stolen profile photo

Almost every fabricated identity needs a face. There are only three places to get one, and each fails in a different, recognisable way.

The three kinds of fake photo

Stolen from a real person. By far the most common. Someone lifts photos from a real, uninvolved person's public accounts — often a model, a soldier, a doctor, or simply an attractive stranger — and reuses them. The victim here is doubled: the person being deceived, and the person whose face is being used, who usually finds out only when strangers start contacting them.

Stock photography. Cheaper and lazier. Recognisable by a quality and composition that doesn't match the story: studio lighting, a neutral background, a pose that exists to sell something, and no ordinary photos alongside it.

AI-generated. Increasingly common, and the category where advice ages fastest. A generated face has never existed, which defeats the single best check available — searching for where else it appears.

Start with reverse image search

It costs nothing and it resolves the first two categories outright.

Save the profile photo, run it through a reverse image search, and read the result as one of three outcomes:

Found under a different name. You are done. If the face belongs to a stock library, an influencer with a different name, or a news article about someone else entirely, the identity in front of you is fabricated. No further investigation required.

Found in a consistent context. Encouraging. The same face on a company team page, a race result, a local newspaper, a conference listing — places the person did not create for your benefit — is genuine corroboration.

Nothing found. Neutral, and this is where people go wrong. Most real people's personal photos are not indexed anywhere. A null result is the normal outcome for an ordinary private person, and it is also what you get for an AI-generated face. It is not evidence in either direction.

Run several photos, not one. A scammer who cropped or mirrored the image they stole may defeat a match on that file while a second photo matches cleanly.

What reverse image search actually does

Worth being precise, because the capability is routinely overstated — including by services selling it.

It matches images, not faces. It is looking for copies and near-copies of the file you gave it: the same photo republished, resized, cropped, mildly edited.

The practical consequences:

• A different photo of the same person usually will not match. There is no "find this human across the internet" button.
• Heavy cropping, mirroring, filters, or re-compression can defeat a match.
• It cannot tell you when a photo was taken, or by whom.
• It cannot tell you whether a matched page is the original source. Search results are not ordered by chronology, and the widely-shared copy often outranks the origin.
• A match on a similar image is not a match on the same image. Two people in similar poses against similar backgrounds will sometimes surface each other.

Treat every hit as a lead to check, never as a verdict.

Tells for a stolen photo

When the search is inconclusive, the photos themselves still carry information:

Quality inconsistent with the life described. A set of professionally lit, well-composed images from someone whose story involves no reason to have them. Real people's photo sets are mixed — good ones, bad ones, ones with other people in them, ones taken by someone else.

Cropping that removes context. Watermarks cut off at the edge, a hand on a shoulder belonging to someone who has been cropped out, an odd aspect ratio. These are signs of a photo taken from somewhere it belonged.

Background that contradicts the story. Signage in a language that doesn't fit. Cars on the wrong side of the road. Plug sockets, licence plates, architecture, or seasons that don't match where they claim to be.

No mundane photos. A genuine account accumulates the unremarkable: bad lighting, food, a pet, a group of friends, a screenshot. A profile that is exclusively flattering portraits is a curated set, and curated sets are often curated from someone else's life.

Nobody else is ever tagged. Real social lives leave two-way traces — other people posting them, tagging them, replying in a way that implies shared history.

Tells for an AI-generated face

Handle these with more caution than most articles suggest, because the reliable tells have been disappearing steadily and any specific list has a short shelf life. As of now the residual artifacts tend to appear at the edges rather than in the face itself:

• accessories that don't resolve — asymmetric earrings, glasses frames that change thickness or don't meet the ear
• teeth and ears rendered with subtly wrong counts or geometry
• hair that dissolves into the background rather than ending
• text anywhere in the image — signs, labels, clothing — rendered as approximations of letters
• backgrounds that fall apart under attention: railings that don't line up, patterns that don't repeat correctly, a second person who is not quite a person
• a face perfectly centred with a shallow, uniform background — a portrait generator's default framing

Do not build your judgement on this list. Build it on the structural check instead, which does not age: a generated face comes with no history. One flawless portrait, or several portraits of the same face in no real place, with no photo of that person at a wedding, on a beach, in a bad group shot, across several years. Generators produce faces easily and lives with difficulty.

The check that outperforms all of the above

Stop analysing pixels and look for time and corroboration.

A real person's images exist across years and across contexts they did not control: tagged by other people, in the background of someone else's photo, on an employer's site, in a club roster, in a local paper. That distribution is extremely hard to fabricate and easy to check.

So the useful question is not "does this photo look fake?" It is: does this face appear anywhere that this person didn't put it? If the entire visual record of someone exists only on the profile talking to you, the photos are the least of what is missing.

Where this fits

Photo checking is one step in confirming someone is real — see how to verify someone before you meet them for the full order, and romance scam warning signs for the pattern these photos are usually attached to.

One limit worth stating plainly: none of this identifies a stranger from a photograph, and it is not meant to. Confirming that someone contacting you is who they claim is reasonable. Using image search to work out who an unknown person is, where they live, or how to reach them is a different activity with a different name, and the fact that a photo was public does not make it acceptable.