How to Tell If a Photo Has Been Manipulated with AI

AI-generated images have reached an astonishing level of realism. In 2026, a photo created with Midjourney, DALL-E, or Stable Diffusion can be indistinguishable from a real photograph to the untrained eye. This poses a serious problem: how do you know if what you’re seeing is real or fabricated? In this article I teach you how to tell if a photo has been manipulated with AI using visual signals and free tools.

Table of contents

Why it’s important to detect AI-manipulated photos

It’s not just about curiosity. Fake photos generate real consequences:

Disinformation: Fake images of events that never happened go viral in minutes, especially during crises or elections. A photo of a disaster that didn’t exist can influence thousands of people’s decisions.

Scams: Fake profiles with AI photos deceive people on dating apps, social media, and sales platforms. Romance scams with AI-generated photos have increased 300% since 2024.

Reputation: Manipulated photos of people can destroy professional and personal reputations. Politicians, celebrities, and ordinary citizens are victims.

Purchase decisions: Fake or AI-enhanced product images lead to disappointing purchases.

I’ve seen friends share “historic” photos that were completely AI-fabricated, and the worst part is nobody questioned their authenticity. The problem isn’t the technology, but the lack of verification.

Pro-tip: Before sharing any striking image on social media, invest 30 seconds in verifying it. Most detection tools are free and take less time than writing a tweet.

Visual signals of an AI-manipulated photo

Although generative AIs keep improving, they still make characteristic mistakes. Here’s what to look for:

Hands and fingers

Still the most common weakness. Generative AIs still struggle with:

• Incorrect number of fingers (6 or 4 instead of 5)
• Fused or deformed fingers
• Joints in impossible positions
• Irregular or oddly angled nails

Text and letters

If the image includes text (signs, posters, clothing brands):

• Deformed or unrecognizable letters
• Words that look like invented languages
• Typographic inconsistency (same font changes between letters)
• Unrecognizable numbers

Facial and body symmetry

AI-generated faces may have:

• Ears different from each other
• Unnatural asymmetry in eyes or eyebrows
• Teeth too perfect or too irregular
• Skin too smooth, doll-like

Backgrounds and details

• Objects in the background that don’t make sense (impossible buildings, doors leading nowhere)
• Blurry transitions where they should be sharp
• Repetitive patterns in textures (grass, hair, fabric)
• Reflections that don’t match the scene

Impossible physics

• Shadows in inconsistent directions
• Reflections not matching subject position
• Incorrect proportions (hands too large, head out of scale)

The best free tools to detect AI photos

After testing dozens of tools, these are the ones that actually work:

Hive Moderation

The most reliable one I’ve tested. Upload an image and it tells you with a confidence percentage whether it was AI-generated.

How to use it:

1. Go to hivemoderation.com.
2. Upload the image or paste the URL.
3. In seconds you get a result: “AI-Generated” or “Not AI-Generated” with percentage.

AI or Not

Simple interface and fast results. Supports images generated by DALL-E, Midjourney, Stable Diffusion, and others.

FotoForensics

Uses ELA (Error Level Analysis) to detect edits. It doesn’t specifically detect AI, but reveals manipulations like cropping, pasting, or color alterations.

Google Reverse Image Search

Search the image on Google to see if it appears in other contexts. If a supposedly real photo appears on stock sites or with different dates, it’s a sign of manipulation.

TinEye

Similar to Google Reverse Search but more specialized. Finds the origin of an image and its different versions on the web.

Pro-tip: Always use two different tools to verify an image. None is 100% reliable on its own. If Hive Moderation says it’s AI and AI or Not says it’s not, investigate more before deciding.

Metadata: the photo’s digital fingerprint

Every digital photo contains EXIF metadata that reveals information about how it was created:

Useful data:

• Camera brand and model
• Capture date and time
• GPS coordinates (if the camera allows it)
• Software used to edit the photo
• Camera settings (ISO, aperture, speed)

How to verify:

1. On Windows: right-click the image > Properties > Details.
2. On Mac: open the image in Preview > Tools > Inspector.
3. In browser: use tools like Jeffrey’s EXIF Viewer.

Alarm signals in metadata:

• Software listed as “Adobe Photoshop” or “Stable Diffusion”
• Complete absence of EXIF data (cameras always generate it)
• Camera listed as “Unknown” or a model that doesn’t exist
• Creation date inconsistent with the image’s context

Note: metadata can be deleted or falsified, so its absence doesn’t prove anything, but its presence can confirm suspicions.

How to act if you discover a fake photo

If you confirm a photo was manipulated or AI-generated:

Don’t share it. Even to debunk it, sharing gives it exposure. Better link to the verification.

Report on the platform. All social media platforms have options to report fake or manipulated content.

Inform whoever shared it. If someone close shared the photo without knowing it’s fake, let them know politely. Many people have no idea this technology exists.

Save evidence. Save screenshots with the date and original source. If the photo is used for scams or defamation, this evidence can be legally useful.

FAQ: Frequently asked questions

Are detection tools 100% reliable?

No. Generation technology advances faster than detection technology. The best tools are correct 85-95% of the time, but there are always false positives and false negatives. Use them as a first verification, not as definitive proof.

Can I detect an AI photo just by looking?

Sometimes, if the image has obvious errors (hands, text, symmetry). But the most recent AI photos are increasingly difficult to visually distinguish. Detection tools are more reliable than the human eye.

Are all photos without metadata fake?

No. Many platforms remove metadata when uploading images for privacy. Its absence doesn’t prove manipulation, but its presence can confirm authenticity.

What if an AI photo affects me personally?

If someone created a fake image with your face, contact the platform to have it removed and consider legal action. In many countries, creating non-consensual images with someone’s face is illegal.

Conclusion

In a world where AI can generate a realistic photo in seconds, visual verification has become an essential skill. You don’t need to be a digital forensics expert: with free tools like Hive Moderation and an eye trained for basic signs, you can detect most fake images. The next step before sharing any striking photo is to verify it. Your credibility and security depend on it.