en flag
en flag

How to Check If an Image Is AI-Generated (2026)

Google's SynthID Detector is now free worldwide. Here's how invisible AI watermarks work, what the checker can and can't tell you, and how to use it.

10/18/20263 min read

Is That Image AI? Google's Free Checker Is Now Open to Everyone, AI Prompt Agent featured graphic
Is That Image AI? Google's Free Checker Is Now Open to Everyone, AI Prompt Agent featured graphic

You've probably seen a photo online and wondered: is this real, or did an AI make it? In October 2026, checking got a lot easier. Google opened its SynthID Detector to everyone, and it doesn't only check Google's own AI. Here's what changed, how invisible watermarks work, and what the tool can't tell you.

New to prompts? Grab our free 10-prompt PDF for ChatGPT, Claude and Gemini.

What happened

On October 7, 2026, Google announced that the SynthID Detector is available globally in English (Google blog). Anyone can now check whether an image, video or audio file was made with AI from Google or its partners, which Google lists as OpenAI, NVIDIA, Kakao and, soon, Apple.

Ars Technica reports that the checker lives on a dedicated website, SynthID.com. Before this, people outside Google's tester program could only check by asking Gemini (Ars Technica).

The surprising number

According to Google, SynthID has been used to watermark over 180 billion images and videos and 240,000 years of audio since it launched in 2023. Google also says its built-in verification features in Search, the Gemini app and Chrome handle over 1 million requests a day.

That's a lot of invisible labels floating around the internet.

How an invisible watermark works

There are two main ways AI content gets "labelled", and they work very differently.

SynthID: a signal inside the pixels

SynthID hides a signal in the image pixels (or in the audio waveform). You can't see or hear it, but a detector can find it. OpenAI's help center explains that because the signal is part of the image itself, it "may persist through some edits or transformations", such as screenshots (OpenAI Help Center).

C2PA: a label attached to the file

C2PA "Content Credentials" are signed metadata that travel with the file and can describe where it came from and how it was edited. It's not only for AI: camera makers and news organizations use the same standard. The weakness, as OpenAI notes, is that metadata can be stripped, for example by uploads, format changes or screenshots.

OpenAI says images made with ChatGPT, Codex and its API now carry both C2PA metadata and a SynthID watermark, so one can back up the other (OpenAI).

What the checker can't tell you

This is the most important part, and the sources are clear about it:

  • "No watermark found" doesn't mean "real." Many AI tools don't use SynthID, and signals can sometimes be removed. A negative result doesn't prove an image is authentic.

  • "AI-made" doesn't mean "false," and "real" doesn't mean "true." OpenAI's verification tool says it only confirms whether an image came from its tools. It doesn't confirm the image is accurate, unedited or shown in the right context.

So treat the detector as one clue, not a verdict. Context still matters: who posted it, where it first appeared, and whether reliable sources report the same thing.

Why this matters if you make AI content

If you create AI images or videos for your brand, this is good news, not a threat. Honest labelling builds trust, and platforms increasingly expect it. A few simple habits:

  1. Label AI visuals clearly, for example "AI-generated example" in the caption or on the video itself.

  2. Don't present AI images as real events, people or customer photos.

  3. Use AI for what it's great at: product mockups, moods, backgrounds and animated versions of your own photos.

That's exactly how we label the demo clips we make for our own social channels.

Make honest, eye-catching AI videos

If you want to turn your own product photos into short, clearly labelled AI clips, our Photo to Video AI Prompt Pack has copy-paste prompts for Kling, Runway, Hailuo and Google Flow.

Get better AI results today

Start with our free 10-prompt PDF, or get the full Photo to Video AI Prompt Pack ($19).

Get the free 10 prompts · See the Photo to Video AI Prompt Pack

FAQ

Is the SynthID Detector free?

Google announced it as available to everyone, globally, in English, from October 7, 2026.

Can it detect images from any AI tool?

No. It detects SynthID watermarks from Google and its partners (OpenAI, NVIDIA, Kakao, and soon Apple). Images from other tools may not carry a compatible watermark.

Can I remove a SynthID watermark?

It's designed to survive common edits like cropping and compression. Trying to remove it to pass AI content off as real is exactly the kind of deception labelling is meant to prevent.

What's the difference between SynthID and C2PA?

SynthID is a hidden signal inside the media. C2PA is signed metadata attached to the file. They work best together.