SynthID Detector homepage with an upload box for images, video and audio files — The SynthID Detector homepage at synthid.com, now open to everyone. Image: Google, via TechCrunch.

Google has opened its AI detector to the public. The SynthID Detector website lets anyone upload an image, a video or an audio clip and check it for Google's hidden AI watermark.

Until now, only journalists, media professionals and researchers on a waitlist could use it. The tool is now available globally in English.

I'm Nitin from the Debugged Pro team. We build websites, apps and AI features for clients in India and the USA, and we handle a lot of media from clients, vendors and users. So here is a plain look at what this tool does, where it falls short and how agencies can use it without trusting it too much.

What Google Launched With Its AI Detector

The news came through Google's official announcement, written by Pushmeet Kohli, VP of Science and Strategic Initiatives at Google DeepMind. TechCrunch also reported the SynthID Detector launch on October 7, 2026.

Here are the main facts from both reports:

Open to everyone: The site lives at synthid.com and is available globally in English.
Three media types: It checks images, video and audio files.
Partner coverage: It can spot media made with AI from Google and its partners OpenAI, NVIDIA and Kakao. Google says Apple support is coming soon.
Scale: Google says it has watermarked over 180 billion images and videos, plus 240,000 years of audio.
Daily use: SynthID checks built into Search, the Gemini app and Chrome now handle more than 1 million requests a day.

Engadget added two useful details. The tool is free, but you must sign in with a Google, OpenAI or Apple account. It will also be one of the first tools able to check images made with Apple Intelligence features like Image Playground, once that support arrives.

Which File Formats Work

TechCrunch listed the supported formats. That matters for agencies, because client files come in every shape.

Images: JPG, JPEG, PNG, BMP, WEBP, AVIF, HEIC, HEIF, TIFF, TIF and GIF
Video: MP4, MOV and WEBM
Audio: WAV, MP3, OGG, FLAC, AAC and M4A

How SynthID Works Behind the AI Detector

SynthID is a watermark that Google first launched in 2023. It hides a signal inside the pixels, frames or sound waves of AI-made media. People cannot see or hear it, but Google's software can read it.

According to TechCrunch, Google's Nano Banana, Veo and Lyria models add SynthID to what they create. So do tools like Gemini, Flow, ProducerAI and Vids.

When you upload a file, the detector scans for that signal. If it finds one, it tells you the media was made or edited with AI and can highlight the parts most likely to carry the watermark. For a video, that means marking the stretches of the clip where it found the signal.

SynthID Detector result screen showing a raccoon video with a SynthID detected message — A sample result: the detector flags a video as made or edited with Google AI and marks where the watermark appears. Image: Google.

Notice the small note in that result: the media "may have been edited further since then." Even a positive result only tells you that AI touched the file at some point. It does not tell you what changed or by how much.

What This AI Detector Cannot Do

This is the part most people will skip, and it is the part that matters most. Google's own homepage says it plainly: this is not a general AI detector. It can only detect media from companies that have adopted SynthID.

Engadget pointed out several more limits:

It misses other watermark systems. Many companies use different standards, so there is no universal checker.
Watermarks can be removed. In some situations, the signal can be stripped from a file.
It does not separate "made" from "edited." A fully AI-generated image and a photo with a small AI edit can return the same result.

TechCrunch also noted that Microsoft and Meta run their own watermarking and checking standards. OpenAI keeps its own site for checking content too. None of these tools is perfect.

The homepage also asks users to upload the highest quality file possible for the best accuracy. In practice, a screenshot of a compressed WhatsApp forward is a weak test file.

Question	Can SynthID Detector answer it?
Was this made with Google, OpenAI, NVIDIA or Kakao AI?	Often yes, if the watermark is intact
Was this made with any AI tool at all?	No, only SynthID partners
Is this file 100% human-made?	No, a missing watermark proves nothing
How much of this file was changed by AI?	No, it only shows where the signal appears
Is this a real camera photo?	No, that needs Content Credentials or other checks

Why It Matters for Agencies in India and the USA

Agencies sit in the middle of a lot of media. We get stock images from clients, product shots from vendors, voice-overs from freelancers and reviews from real users. Any of it could be AI-made, and the client usually expects us to know.

A free, easy check changes the daily work in a few ways:

Faster first-pass checks: A content editor can test a doubtful image in under a minute.
Better client conversations: You can show a result instead of saying "it looks fake to me."
Safer user-generated content: Review sites, contests and community apps can screen uploads before they go live.
Clearer disclosure: If your own team uses Gemini or ChatGPT image tools, you can confirm which assets carry a watermark before you label them.

We wrote earlier about AI watermarking rules and what they mean for agency work. This launch is the practical side of that story. The watermark only helps if someone actually checks it, and now anyone can.

For design-heavy teams, a detector is only one layer. Our guide on how to spot AI-generated design covers the visual signs that a tool will not flag.

How to Use an AI Detector in Your Content Workflow

Here is a simple process we would suggest for any agency that handles client media. It keeps the tool in its proper place: useful, but not the final word.

Decide what needs checking. Focus on risky items: news-style photos, testimonials, "real customer" images, audio quotes and user uploads.
Ask for the original file. Request the highest quality version, not a screenshot or a compressed copy.
Run the SynthID check. Upload the file at synthid.com and save a screenshot of the result.
Check Content Credentials too. Many tools now attach C2PA Content Credentials, an open standard that records how a file was made or edited.
Read a negative result as "unknown." No watermark does not mean human-made. Confirm the source with the client or vendor.
Log the outcome. Keep a short note with the file name, date, result and who checked it.

A Real-World Example

Say a client in Bengaluru sends 20 "customer photos" for a new landing page. Your editor runs them through the detector. Three come back with a SynthID match.

That does not mean the client lied. Maybe a junior marketer used Gemini to clean up the backgrounds. The right move is a quick call, a question about how the images were made and a decision on whether to label them or swap them out.

The same check helps a US client running a review site. A flagged upload can go to a human moderator instead of straight to the live page.

Building AI Detector Checks Into Your Products

Some clients will want this inside their own apps, not as a manual step. Google has a path for that. In a May 2026 post, Google said it was launching an AI Content Detection API on Google Cloud's Gemini Enterprise Agent Platform, starting with a group of trusted partners.

Google said that API can spot content from both Google and other popular models. It pointed to uses like sorting feeds, preventing insurance fraud, fact-checking and labeling synthetic media. Access was limited at launch, so check availability before you promise it to a client.

If you are planning this kind of feature, think about these questions early:

What happens when a check fails or times out?
Who reviews flagged items, and how fast?
How do you explain a "likely AI" label to end users?
Where are uploaded files stored, and for how long?

Our software development consulting team helps clients plan features like content moderation queues and media checks, so the tech choice fits the actual risk.

Chat prompt asking Is this generated with AI next to a photo of a dog in a coffee cup — Google also lets people ask Gemini and Search whether an image is AI-generated. Image: Google.

Our Take on Google's AI Detector

This is a useful step, and the price is right. Making detection free and simple will get more people in the habit of checking before they share or publish.

But agencies should be careful not to oversell it. The tool answers one narrow question: does this file carry a SynthID watermark? It cannot clear a file as human-made, and it cannot catch content from tools outside the SynthID group.

Coverage should grow. Engadget noted that ElevenLabs had committed to adding SynthID support, and Google says it is working to add more partners. For now, treat it as one good test in a set of checks, not a lie detector.

Key Takeaways

Google's SynthID Detector is now open to everyone at synthid.com, free with a Google, OpenAI or Apple sign-in.
It checks images, video and audio for SynthID watermarks from Google, OpenAI, NVIDIA and Kakao, with Apple support coming.
It is not a general AI detector. A missing watermark proves nothing.
Agencies should use it for risky assets, ask for original files and log each result.
Pair it with Content Credentials checks and plain human review for the best results.

Frequently Asked Questions

Is Google's SynthID Detector free to use?

Yes. Engadget reports that the tool is free, but you need to sign in with a Google, OpenAI or Apple account before you can upload files at synthid.com.

Can SynthID Detector tell if any image was made with AI?

No. It only finds SynthID watermarks. Google's own page says it is not a general AI detector and can only detect media from companies that have adopted SynthID, such as Google, OpenAI, NVIDIA and Kakao.

What file types does the SynthID Detector accept?

TechCrunch lists JPG, JPEG, PNG, BMP, WEBP, AVIF, HEIC, HEIF, TIFF, TIF and GIF for images, MP4, MOV and WEBM for video, and WAV, MP3, OGG, FLAC, AAC and M4A for audio.

If no watermark is found, does that mean the content is human-made?

Not at all. The file may come from a tool that does not use SynthID, or the watermark may have been stripped. Treat a negative result as unknown and check the source in other ways.

Should agencies run every client asset through an AI detector?

Not every asset. Focus on high-risk items like news-style images, testimonials, user-submitted media and anything a client will use as proof. Log each check so you can show your process later.

About the Author

Nitin Kumar is the Lead Project Coordinator at Debugged Pro. Nitin works with clients in India and the USA on web, mobile and AI projects, turning new tech news into practical plans that teams can actually use.