mekat.
Try a lesson

How to tell if an image is AI-generated

You often can’t tell by looking anymore. AI images have become very realistic, so the reliable approach is a mix of three checks: look for clues, check for a watermark or Content Credentials, and, most important, check where the image came from. A missing label never proves an image is real.Updated September 29, 2026

Start with the source, not the pixels

Before you zoom in on fingers, ask the questions that work for any photo:

  1. Who posted it first? An account you can’t trace, a brand-new profile or a page that only posts viral images is a warning sign.
  2. Is anyone reliable reporting the same thing? A real photo of a major event usually appears in several trusted news outlets, taken from different angles.
  3. Does it push you to feel something fast? Outrage, fear or “share before they delete this” is how fakes spread.
  4. Run a reverse image search (Google Lens, Bing Visual Search or TinEye). It can show where the image appeared before, including earlier versions or the original it was edited from.

Visual clues that still help

AI models keep improving, so treat these as hints, not proof. The FBI lists subtle imperfections that can give AI images and videos away:

  • Hands or feet that are distorted or have the wrong number of fingers or toes.
  • Teeth or eyes that look unrealistic.
  • Faces that are blurry, smeared or irregular, especially people in the background.
  • Glasses, earrings and other jewellery that don’t look right.
  • Shadows that fall the wrong way.

Other common giveaways: writing on signs, labels or posters that turns into nonsense up close, and skin, hair or backgrounds that look too smooth, like a painting.

In videos, the FBI also points to lag, voices that don’t match and unrealistic movement.

Check for a watermark or Content Credentials

Some AI tools now mark what they make, and free tools can read those marks.

  • SynthID. Google DeepMind’s SynthID adds an invisible watermark to images, video, audio and text made with Google’s AI tools, designed to survive cropping, filters and compression. In the Gemini app you can upload an image and ask whether it was made with Google AI; Google says it can tell you if a SynthID watermark is found. Apple says Apple Intelligence image tools will add SynthID in a software update later in 2026.
  • Content Credentials (C2PA). An open standard that works like a nutrition label for images: it can record which camera or app made a picture and how it was edited. Upload an image to the free Content Credentials Verify site to see its history, if it has one.

The catch: each check only finds its own marks. “No SynthID found” means Google’s AI didn’t make it, not that no AI did. Many AI tools add no mark at all, and screenshots or re-uploads can strip Content Credentials. So a mark is strong evidence, but no mark isn’t evidence of anything.

What about AI detector websites?

Sites that give any image an “AI score” are tempting, but they can be wrong in both directions. NIST’s 2024 report on synthetic content notes that detectors often work well only on the specific AI generators they were built for, and that false positives, calling a real photo AI-made, can do serious harm to the people involved. Use a detector as one clue among several, never as the final word, and never to accuse someone of faking a photo.

When it matters, slow down

Most AI images are harmless fun. The ones worth checking are those that ask you to believe, buy or share something: a disaster photo, a celebrity “endorsing” an investment, a picture of a person that could hurt their reputation, or a profile photo on someone asking for money. For those, don’t share it until you’ve checked the source, and report scams to the FTC at ReportFraud.ftc.gov.

How Mekat helps

Mekat’s lessons on images, audio and documents and on privacy, safety and scams teach you to spot deepfakes and check sources, with practice on realistic examples. See the full learning path.

Quick answers

Can you always tell if an image is AI-generated?

No. The best AI images can fool experts. Combine visual clues, a watermark or Content Credentials check and, above all, a check of where the image came from.

How do I check if an image was made with Google AI?

Upload it to the Gemini app and ask if it was made with Google AI. Gemini checks for Google’s invisible SynthID watermark. It can’t detect images from other companies’ tools.

Are AI image detectors accurate?

Not reliably. They can flag real photos as fake and miss AI images. Use them as one clue, alongside the source and any Content Credentials.

What are Content Credentials?

An open standard (C2PA) that attaches a record of how an image was made and edited. You can check an image for free on the Content Credentials Verify site.

Sources

Written by the Mekat team at Martinez Solutions LLC, Glendale, Arizona. Spotted something out of date? support@martinezsolutionsaz.com