# SynthID Detector Is Now Public: How to Check If Media Was Made With AI

- Link: https://thejusgrow.com/synthid-detector/
- Published: 2026-10-11T19:34:26+00:00
- Author: Varun Sharma

SynthID Detector in brief

The **SynthID Detector** checks images, video and audio for SynthID, an invisible
watermark that Google and some partners build into AI-generated content. Since 7
October 2026 anyone can use it, worldwide and in English, at synthid.com. A positive
result is strong evidence a file came from a participating AI tool. A negative result
proves nothing about whether the file is real.

Google opened the SynthID Detector to the public on 7 October 2026, and it’s genuinely
useful. It’s also going to be misunderstood by a lot of people, because the name
suggests something it isn’t. This is not a general “is this AI?” button. It looks
for one specific kind of hidden watermark, and if that watermark isn’t there, it
has nothing to tell you.

That distinction is the whole point of this article. Used properly, the SynthID 
Detector is a fast way to confirm that a suspicious image or voice clip came from
an AI tool. Used carelessly, a “nothing found” result can make a fake look more 
trustworthy than it deserves.

What follows draws on Google DeepMind’s [announcement](https://blog.google/innovation-and-ai/models-and-research/google-deepmind/synth-id-ai-content/),
its [SynthID product page](https://deepmind.google/models/synthid/) and Google’s
earlier posts about the tool, all read in October 2026.

## Watermarks and detectors are different things

There are two broad ways to spot AI-generated media, and it helps to separate them.

The first is a **classifier**: software trained to look at any image or recording
and guess whether a machine made it. Classifiers work on anything, but they guess,
and they get things wrong in both directions.

The second is a **watermark**. The company that makes the AI tool hides a signal
in everything the tool produces, and a matching detector looks for that signal later.
Watermark detection is far more reliable when the signal is there. The catch is 
obvious: it only works on content from tools that add the watermark in the first
place.

SynthID is the second kind. According to Google DeepMind, the watermark goes in 
at the moment of creation, can’t be seen or heard, and is designed not to affect
quality. The SynthID Detector simply checks whether that mark is present.

## What changed on 7 October

The detector itself isn’t new. Google showed an early version at its I/O conference
in May 2025, but access was limited to testers and a waitlist of journalists, media
professionals and researchers. Most people never got to use it.

Now anyone can. Google says it is available globally, in English, and handles images,
video and audio. The bigger change is whose content it can recognise. Alongside 
Google’s own tools, the detector can now check content from **OpenAI, NVIDIA and
Kakao**, with **Apple** listed as coming soon.

That partner list is what makes the public launch matter. A watermark detector that
only recognised Google’s images would have limited value when so much AI content
comes from elsewhere. With OpenAI on board, and Apple on the way, a much bigger 
share of everyday AI media becomes checkable.

Google also shared some scale figures: more than 180 billion images and videos and
240,000 years of audio watermarked since SynthID launched in 2023. It says verification
features built into Search, the Gemini app and Chrome now handle more than a million
requests a day.

## How to use the SynthID Detector

Google’s description of the process is short. You go to synthid.com, the portal 
linked from Google DeepMind’s SynthID page, upload an image, video or audio file,
and the tool reports whether it found a SynthID watermark.

When Google demonstrated the early version in 2025, it said the results screen highlighted
the parts of a file most likely to carry the watermark, down to specific segments
of an audio clip. The October 2026 announcement doesn’t describe the current results
screen, so it may look different now.

There’s a second route that many people will find easier. Google says you can upload
a file to Gemini, Search or Chrome and simply ask whether Google AI made or changed
it. When Google added this to the Gemini app in November 2025, it suggested questions
such as “Was this created with Google AI?” or “Is this AI-generated?”.

## Reading a SynthID Detector result correctly

This is where most mistakes will happen, so it’s worth being precise.

**Watermark found.** This is strong evidence that the file was made or edited with
a tool from Google or one of its SynthID partners. It doesn’t tell you who made 
it, why, or whether it’s meant to deceive anyone. Plenty of watermarked images are
harmless illustrations.

**No watermark found.** This only means the detector didn’t find a SynthID signal.
The file could be a real photo. It could equally come from an AI tool that doesn’t
use SynthID, of which there are many. It could even have been altered heavily enough
to damage the mark.

There’s a middle case too. Google describes these checks as telling you whether 
its AI made or changed a file. A real photograph that someone has touched up with
a Google AI editing tool may therefore carry the watermark, even though most of 
what you see is genuine. A positive result tells you AI played a part. It doesn’t
tell you how big that part was.

If you remember one thing, make it this: a missing watermark means “unknown”, not“
real”. Google doesn’t claim otherwise. Its tools report whether they find a SynthID
watermark, and nothing more. Google also hasn’t published accuracy or error rates
for the detector, which would help people judge how much weight to put on a result.

## How hard is the watermark to remove?

Google says the image and video watermark is “designed to stand up to” cropping,
filters, frame-rate changes and lossy compression. For audio, it says common changes
such as added noise, MP3 compression or speed changes don’t remove it.

Those claims matter, because social media platforms routinely crop and recompress
everything uploaded to them. If the watermark survives that journey, the detector
stays useful for files that have been shared and reshared.

They’re still Google’s claims, not independent results, and no watermark scheme 
is unbreakable. A determined person with technical skill may find ways to weaken
it. For everyday misinformation, which usually travels through screenshots and reposts
rather than deliberate attacks on watermarks, the protection Google describes is
likely to hold up better than for a well-resourced faker.

SynthID can also watermark text from the Gemini app and web experience, by subtly
shaping which words the model picks. The October announcement focuses on images,
video and audio, though, and that’s where the public detector is aimed.

## Where the SynthID Detector helps

A few everyday situations where it’s worth reaching for:

 * **A viral image that seems too perfect.** Before sharing a dramatic photo of 
   a news event, a quick check can confirm whether a participating AI tool generated
   it.
 * **A voice note that sounds like a relative asking for money.** If the detector
   finds a watermark, you have your answer. If it doesn’t, that proves nothing, 
   and the safer move is a call back on a number you already know. Our guide to 
   [family code words for AI voice scams](https://thejusgrow.com/family-code-word-ai-voice-scams/)
   explains a simple defence that works even when no tool can help.
 * **Images sent to a business.** Small shops and insurers increasingly receive 
   photos as evidence of damage or delivery. A watermark check is a cheap first 
   step before trusting them.
 * **Classrooms and newsrooms**, where people need to check media regularly and 
   explain how they did it.

Scammers rarely announce which tools they used, so in most of these cases the detector
is one check among several. Our article on [AI agent scams](https://thejusgrow.com/ai-agent-scams/)
covers other tricks that rely on convincing fake media.

## The limits worth knowing

Based on what Google has published, the main limitations are:

 * It only finds SynthID watermarks, so content from tools that don’t use SynthID
   won’t register.
 * It’s English-only at launch.
 * Apple support isn’t live yet.
 * There are no published accuracy or error rates.
 * A result can confirm AI involvement but can’t judge intent or truthfulness.

Google has also said, in its November 2025 post about the Gemini app, that it plans
to support C2PA content credentials over time. C2PA is an industry standard for 
attaching provenance information to files. If that happens, checks inside Google’s
products could cover content from a much wider range of tools than SynthID alone.

## Should you bother with the SynthID Detector?

Yes, as long as you treat it as a quick test rather than a verdict. It takes seconds,
it’s free, and for content from Google, OpenAI, NVIDIA or Kakao tools it can give
a clear answer. That’s a real improvement on a year ago, when most people had no
way to check at all.

What it can’t do is tell you a picture is genuine. For that, the old habits still
matter: find the original source, check whether trusted outlets are reporting the
same thing, and be most suspicious of content designed to make you angry or rush
you into sending money. Our guide on [how to spot a deepfake video](https://thejusgrow.com/how-to-spot-a-deepfake-video/)
walks through those checks.

## SynthID Detector: FAQs

**Is the SynthID Detector free?**

Google offers it as a public tool at synthid.com and its announcement mentions no
charge. It became available globally, in English, on 7 October 2026.

**Can the SynthID Detector spot any AI-generated image?**

No. It only detects SynthID watermarks, used by Google and partners including OpenAI,
NVIDIA and Kakao. Images from other AI tools won’t show a watermark.

**If no watermark is found, is the image real?**

Not necessarily. The file could come from an AI tool that doesn’t use SynthID, so
check its original source too.

## Sources

 * [Google DeepMind, “Google expands SynthID Detector for AI content” (7 October 2026)](https://blog.google/innovation-and-ai/models-and-research/google-deepmind/synth-id-ai-content/)
 * [SynthID product page, Google DeepMind (accessed October 2026)](https://deepmind.google/models/synthid/)
 * [Google, “SynthID Detector: a new portal to help identify AI-generated content” (20 May 2025)](https://blog.google/innovation-and-ai/products/google-synthid-ai-content-detector/)
 * [Gemini app announcement, “How we’re bringing AI image verification to the Gemini app” (20 November 2025)](https://blog.google/innovation-and-ai/products/ai-image-verification-gemini-app/)