How to Tell If an Image Is AI-Generated: 6 Checks

AI image generators can now produce faces, products and news-style scenes that look like real photos. Face-swap tools can put one person's face onto another person's body in a real photo. Neither is always easy to catch by eye, but a few checks, done in the right order, will catch many fakes and help you decide how much to trust a picture before you share it or act on it.

AI-generated images vs face-swap deepfakes

It helps to know which kind of fake you might be dealing with, because the signs differ.

• AI-generated image: created from scratch by an image model from a text prompt. Nothing in it was ever photographed. Fake profile pictures, invented events and fake product photos often fall into this group.
• Face-swap deepfake: starts from a real photo and replaces the face with someone else's. The background, body and clothes are real, which makes these harder to spot.

For a broader introduction, see our guide on what a deepfake is and how deepfakes are made.

Step 1: Check the source before the pixels

The fastest check usually has nothing to do with the image itself.

• Who posted it first? A new account, an anonymous page or a repost with no credit is a reason to slow down.
• Do trusted sources show the same thing? If an image shows a major event, reputable news outlets will usually have their own photos of it.
• Is it pushing you to act now? Urgency, outrage and "share before it gets deleted" are common in manipulated content.

Step 2: Run a reverse image search

Upload the image to a reverse image search such as Google Lens, Bing Visual Search or TinEye. This can show you:

• an older or original version of the photo, which reveals a face swap or a reused picture;
• the first place it appeared, which helps you judge the source;
• fact-checks that have already covered it.

No results does not prove an image is fake, but a brand-new "photo" of a big event with no other copies anywhere is worth questioning.

Step 3: Look for visual signs of AI generation

Zoom in and check the areas where image models still tend to slip:

• Hands and fingers: extra, missing or merged fingers, and hands that bend in odd ways.
• Text and logos: signs, labels and badges with garbled or nonsense lettering.
• Edges and accessories: earrings that don't match, glasses that melt into the face, hair that blends into the background.
• Teeth and eyes: teeth that run together, pupils of different shapes, or reflections in each eye that don't match.
• Background: warped lines, repeating patterns, people in the background with distorted faces.
• Overall look: skin that is too smooth, lighting that is too perfect, or a glossy look that resembles digital art more than a camera photo.

Newer generators make fewer of these mistakes, so a clean result on this checklist is not proof that an image is real.

Step 4: Look for signs of a face swap

Because a face swap pastes a face into a real photo, look at where the two meet:

• a visible seam, blur or colour change around the jawline, hairline or ears;
• lighting and shadows on the face that don't match the rest of the scene;
• a face that is sharper or softer than the neck and body;
• skin tone that changes between the face and the neck or hands.

Step 5: Check for content credentials and metadata

Some cameras, editing apps and AI tools attach Content Credentials based on the C2PA standard, which record how an image was made or edited. You can inspect them with the free Content Credentials verify tool. Some AI generators also add a label or invisible watermark.

Treat this as a bonus check only. Most images online have no credentials, and social platforms often strip metadata when a picture is uploaded, so missing information tells you very little.

Step 6: Use a deepfake image detector

Detection models look for statistical traces that AI generators and face-swap tools leave behind, including some that are invisible to the eye. DeepWall's free deepfake image detection tool lets you upload a JPG, PNG or WEBP image up to 4 MB with no sign-up. It returns two separate scores: one for AI generation and one for face swapping, plus the AI generator the image most likely came from when that is available.

Keep the limits in mind. Scores are probabilities, not proof. Heavy compression, very small faces, low resolution and screenshots of screenshots all reduce accuracy, and the tool is not built to detect ordinary edits such as cropping, filters or retouching. Use the score as one signal alongside the source, reverse image search and visual checks above. To check a video clip or an audio file as well, use the free deepfake detector, which puts the image, video and audio checkers on one page.

What to do if you think an image is fake

• Don't share it, even to warn others, without making clear it is suspected to be fake.
• Save evidence: take screenshots and note the URL, account name and date.
• Report it to the platform using its impersonation or manipulated media reporting option.
• If it shows you or your organisation, warn the people it might be used to fool, such as colleagues, customers or family.

Images are only part of the picture

A fake image is often a sign that fake videos or cloned voices of the same person are also circulating. For video, read how to spot a deepfake video or test a clip with our deepfake video detection checker. For voice notes and calls, see our guide to voice cloning scams and our deepfake audio detection page.

If you need to know when your face or voice is used without permission, DeepWall's deepfake detection software monitors newly uploaded videos for the faces and voices you enroll, so you can act on impersonation you haven't seen yet. Start your free trial.

Frequently asked questions

How can you tell if an image is AI-generated?

Check the source first, then run a reverse image search. Zoom in on hands, text, teeth, eyes, accessories and the background for distortions, and use a deepfake image detector as a final signal. No single check is conclusive on its own.

Can AI-generated images be detected for free?

Yes. DeepWall's free image checker scores an image for AI generation and face swapping with no sign-up, up to 5 images per visitor per day. Reverse image search tools and the Content Credentials verify tool are also free.

Are AI image detectors always accurate?

No. They catch many fakes that look convincing to the eye, but accuracy drops with compression, small faces and heavy edits. Treat a score as one piece of evidence, not proof.

What is the difference between an AI-generated image and a deepfake?

An AI-generated image is made entirely by a model. A face-swap deepfake starts from a real photo and replaces the face. People often use "deepfake" for both.

Does missing metadata mean an image is fake?

No. Most social platforms strip metadata on upload, so most real photos online have none. Metadata or Content Credentials can help when they are present, but their absence proves nothing.