Is That a Real Face — or a Deepfake?

Deepfakes aren't just a celebrity problem anymore

Deepfake fraud is no longer a niche problem. Sumsub's 2024 identity fraud report found deepfake incidents grew 257% in a single year — and most of that growth wasn't celebrity face-swaps. It was profile photos on dating apps, recruiter accounts on LinkedIn, reviewer headshots on product listings. The face looks convincing because the generation tools have gotten good enough to produce something indistinguishable from a real photograph.

The photos aren't always taken by a camera anymore. Some were assembled from scratch by a model that never looked at a real person. And nothing about them signals this to the eye.

How do you detect a deepfake?

1. Install Faux Spy in Chrome, Firefox or Edge (free, 3 checks per day)
2. Open any website with the photo you want to check
3. Hover over the image — the Investigate button appears
4. Click it. You'll have a verdict and confidence score in seconds

No uploading images to a third-party site. No copying URLs. The check runs directly from your browser on the page you're already on.

What do deepfake detection results mean?

AI Photo — this is the deepfake category. The model detected a photorealistic AI-generated image: a synthetic face that doesn't belong to any real person. Above 80% confidence, take it seriously.

Real Photo — the image has the characteristics of an authentic photograph. A 95% "Real Photo" verdict is strong. A 55% one is worth noting but not conclusive.

Inconclusive — not enough signal. Usually means the image is too small or compressed. Find the full-size version and try again.

Where do deepfakes show up?

Dating apps and social media — fake profile photos are the most common use case. Generated faces tend to have a specific quality: everything in proportion, skin unusually clean, lighting flattering in a way candid photos rarely are. Real faces have more randomness. Your brain notices something without being able to name it.

LinkedIn and professional networks — fake recruiter profiles, phantom candidates, and scam outreach accounts often use generated headshots because they look more professional than stolen real photos. A polished headshot is less suspicious than a photo grabbed from someone's Instagram.

News and disinformation — generated images attached to real events: "witnesses" at an incident, protest crowds, public figures in places they never were. The image makes something look photographed that no camera ever captured.

Product reviews and testimonials — review fraud increasingly uses AI-generated reviewer photos. A cluster of reviewer headshots with the same studio-quality lighting, no glasses, no specific background, nothing personal — no real customer photographs like that.

How does a deepfake detector work?

Image generators leave statistical fingerprints in the pixels — in textures, noise and the way fine detail is rendered — that a camera doesn't produce. A detection model is trained on large sets of real and AI-generated images until it can score a new image on how closely it matches each.

Faux Spy runs each scan through two detection models, from Hive and Sightengine, and also checks for camera provenance: C2PA Content Credentials and EXIF metadata that can show a photo came from a real camera. The result is a verdict plus a confidence percentage, not a flat yes or no. For a deeper explanation, see how AI image detection works.

Can you spot a deepfake by eye?

Sometimes, with older or lower-quality generators. Look for:

• Hands with the wrong number of fingers, or fingers that merge
• Earrings, glasses or collars that don't match on each side
• Text, logos or signs that turn into gibberish
• Hair that blends into the background, or a background that's oddly smeared
• Skin that's too smooth and lighting that's too flattering for a casual photo

The newest generators fix most of these, so a photo with none of these flaws can still be AI-generated. Treat a visual check as a first pass and use a detector to confirm. And because detectors can be wrong too, a real photo is occasionally flagged — see why a real photo might be flagged as AI.

Deepfake detector vs. other ways to check a photo

Which AI generators can it detect?

Faux Spy detects photorealistic images from the major generators, including Midjourney, DALL·E and ChatGPT image generation, Stable Diffusion, Adobe Firefly and Flux. Generator-specific guides: Midjourney · ChatGPT · DALL·E · Stable Diffusion · Adobe Firefly · Flux. For video from Sora, Runway and others, see the AI video detector.