Guides

How to Tell If a Photo Is Real or AI-Generated in 2026

AI images are now good enough to fool most people at a glance. Studies that ask people to sort real photos from generated ones find results close to guessing. So looking harder is not enough on its own. You need a method.

This guide gives you seven checks you can do in a few minutes with free tools. None of them is perfect. Together they catch most fakes and, just as important, stop you from calling a real photo fake.

1. Start with the source

Before you look at pixels, look at who posted the photo. This is the fastest check and often the most useful one.

• Who posted it first? A news agency, a known photographer or the person who was there is a better start than an account that only reposts.
• When and where? Does the date fit the event? Does the weather, the season or the language on signs fit the place?
• Does anyone else have it? A real event usually has more than one photo from more than one angle. One perfect image and nothing else is a warning sign.

2. Run a reverse image search

A reverse image search looks for the same or similar images on the web. Free options include Google Lens, Bing Visual Search and TinEye. TinEye can sort results by date, which helps you find the oldest copy.

This check is good at two things. It finds the original source, and it shows when an old photo is being shared as new. It cannot prove how an image was made. No match does not mean fake either: a new photo simply may not be online anywhere else yet.

3. Check the metadata

Photos from a camera or phone usually carry metadata, often called EXIF: the camera model, the date and time, settings like exposure, and sometimes the location. Many free online viewers will show it, and most computers show it in the file details.

Know the limits before you trust it:

• Many social apps remove most metadata when a photo is uploaded, partly to protect location privacy.
• Metadata can be edited with free tools.
• So missing metadata proves nothing, and present metadata is a clue, not proof.

4. Look for Content Credentials

Content Credentials are a newer kind of information attached to a file. They follow an open standard called C2PA, backed by a group of camera makers, software companies and media organisations. A file with Content Credentials can show which device or app made it and how it was edited.

Some cameras and phones now add them when you take a photo, and some AI image tools add them to the images they create. You can check a file with a free Content Credentials viewer, such as the one at contentcredentials.org/verify.

The catch is that the credentials travel inside the file. A screenshot, a re-upload or a platform that strips data on upload can remove them. So credentials that are present are useful evidence. Credentials that are missing tell you very little.

5. Look closely at the details

Zoom in. AI images still slip in some places, though less every year:

• Text: signs, labels, screens and number plates with letters that melt or make no sense.
• Hands and ears: extra or merged fingers, earrings that do not match.
• Light: shadows that point different ways, reflections that do not match the thing they reflect.
• Edges and patterns: hair that blends into the background, tiles or bricks that repeat in odd ways.
• Skin: very smooth, even skin with no pores in a photo that is otherwise sharp.

Be careful with this list. Newer generators make fewer of these mistakes, and real photos can show some of them too, because of phone processing, motion blur or editing. A visual sign is a reason to check more, not a final answer.

6. Use AI detectors carefully

Many websites promise to tell you if an image is AI. They give a score or a percentage. Use that score as one clue only.

• Detectors can flag real photos as AI. Ordinary steps like compression, resizing, heavy editing or noise reduction can confuse them.
• Detectors can miss images from newer generators that they were not trained on.
• Two detectors often disagree about the same image.

Never accuse a person of faking a photo because a detector said so.

7. A two-minute checklist

1. Who posted it first, and when?
2. Reverse image search: is there an older copy or another source?
3. Content Credentials: is there a record of how it was made?
4. Metadata: does the date and device fit the story?
5. Details: text, hands, shadows, reflections.
6. Detector: one more clue, not a verdict.
7. Still unsure? Say "not confirmed" and do not share it as fact.

How Rawly handles it

Rawly is a photo and video challenge app. We do not try to detect AI images. We made the question easier to answer inside our own app instead.

• Live camera only. Every photo and video is taken with the camera inside the app. There is no gallery upload, so a finished or generated image has no way in.
• A code on every shared photo. When someone shares or saves a photo from Rawly, it carries a short code that starts with RWLY.
• Anyone can check it. Type the code at rawly.app/verify. You see Rawly's own copy of the photo and the minute it was taken. No account is needed.

It is important to be clear about the limits. The check proves when a photo was taken in Rawly, not who took it. If the photo in front of you looks different from Rawly's copy, someone copied the code onto another image. And it only works for photos taken in Rawly. It says nothing about photos from other apps.

If you want a feed where every photo was taken live, see AI-free social media apps and a social network with no bots.

Questions

How can I tell if a photo is AI-generated?

No single check is enough. Start with the source and a reverse image search, then look for Content Credentials, read the metadata if it is still there, and look closely at text, hands and shadows. If the checks disagree, treat the photo as unconfirmed.

Are AI image detectors reliable?

Not fully. Detectors give a probability, not a verdict. They can flag real photos as AI, especially after compression or editing, and they can miss images from newer generators. Use a detector as one clue among several.

What are Content Credentials?

Content Credentials are information attached to a file under an open standard called C2PA. They can record which camera or app made the file and how it was edited. Some cameras and some AI tools add them. They are easy to lose: a screenshot or a re-upload often removes them.

Does missing metadata mean a photo is fake?

No. Many social apps remove metadata when you upload a photo, and anyone can edit it. Missing metadata tells you very little. Metadata that is present can help, but it can also be changed.

Can a reverse image search prove a photo is real?

No. It shows where else the image appears and, with some tools, when it first appeared. That helps you find the original source or spot an old photo shared as new. It cannot prove how the image was made.

How does Rawly check that a photo is real?

Rawly only accepts photos and videos taken live with the camera inside the app. There is no gallery upload. Every photo shared from Rawly carries a short code, and anyone can type it at rawly.app/verify to see Rawly's copy and the minute it was taken. This only works for photos taken in Rawly.

Can Rawly tell if a photo from another app is AI?

No. Rawly does not scan photos from other apps. The code check at rawly.app/verify only answers one question: was this photo taken in Rawly, and when.