The quickest way to find out whether a suspicious photo, video or audio clip came from an AI tool is Google’s SynthID Detector, a standalone website at synthid.com that is now open to everyone instead of a short list of journalists. The SynthID Detector looks for an invisible watermark that Google and several partner companies stamp into the media their AI tools produce.
That partner list is the part that matters most for families. Google says the portal can flag files made with AI from Google, OpenAI, NVIDIA and Kakao, with Apple coming soon. So the viral picture in the family group chat now has a real chance of getting a straight answer.
What the SynthID Detector actually looks for
SynthID is a watermark you cannot see or hear. When a supported AI model creates an image or a video, a hidden signal gets baked into the pixels at the moment of creation. Google says it does not lower the quality, and it is built to hold up through cropping, filters, frame-rate changes and the heavy compression that happens when a file bounces around messaging apps.
Audio gets the same treatment. Music from Google’s Lyria model and the AI podcast feature in NotebookLM carry an inaudible mark that is designed to resist background noise, MP3 compression and speed changes. Google also marks text written by the Gemini app, though more on that limit below.
The detector simply reads that hidden signal back. It does not guess from how a picture looks, so its answers beat visual hunches. You can read Google’s own overview on the SynthID page from Google DeepMind.
Your two options for running a check
Google now offers two doors into the same technology. Pick whichever one fits the device in your hand.
| Option | Where to find it | What it can check | Best for |
|---|---|---|---|
| SynthID Detector portal | synthid.com in any browser | Images, video and audio from Google AI and partners (OpenAI, NVIDIA, Kakao, Apple soon) | Files that may have come from a non-Google AI tool |
| Gemini verification | gemini.google.com and the Gemini Android app (also built into Search and Chrome) | Images, video and audio, plus Content Credentials history | Quick checks while you are already signed in to Google |
The bigger partner list is the main reason to start with the SynthID Detector website. Google has not published a detailed walkthrough of the portal itself, so the hands-on steps below cover Gemini, where the process is documented.
How to check a photo, video or voice clip in Gemini
According to Google’s Gemini help page, you need to be signed in with a personal Google account or a qualified Workspace account. If you use the phone app, update Gemini to the newest version first.
- On a computer, open gemini.google.com. On Android, open the Gemini app.
- Click Add files and choose a file saved on your computer or phone, in Google Photos or in Google Drive. Drive uploads only work if Keep Activity is turned on and Google Workspace is connected to Gemini Apps.
- Type a plain question, such as “Is this AI-generated?” or “Was this video edited with Google AI?”
- Read the reply. Gemini tells you whether it found a SynthID mark and, for video, which portions carry it.
Checking a screenshot? Crop it tightly so only the picture in question remains. Collages and busy screenshots of a whole social feed tend to confuse the check.
Gemini also reads Content Credentials, an industry standard that records where a file came from and how it was edited. That check works on the web and on Android today, and Google lists iPhone support as coming soon.
How to read SynthID Detector results
The answers are more nuanced than a simple yes or no. Here is what each one means:
- Watermark found. Some or all of the file was created or altered by a supported AI model. A real photo with an AI-edited sky can trigger this too.
- No watermark found. A Google model probably did not make or edit it. Another AI tool that does not use SynthID still could have.
- Unclear. Very plain or abstract images may hold too little detail to carry a mark, and tiny edits may be too small to register.
- Content Credentials present. Gemini sums up how the file was put together, its edit history and whether AI played a part.
A “not found” result is the one most people misread. It does not certify a photo as real. It only tells you that none of the participating AI makers left their fingerprint on it.
Limits worth knowing before you rely on it
Gemini’s checker handles one file at a time, up to 100 MB. Videos must run shorter than 90 seconds and audio shorter than one hour. Within any rolling 24-hour window you get roughly 10 image checks, 10 video checks (5 minutes of footage in total) and 10 audio checks (3 hours in total).
- Heavy editing can erase the trail. The mark usually survives resizing, recoloring and compression, but a file that has been altered over and over may no longer read correctly.
- Coverage depends on who joins. Content from AI makers outside the SynthID group will not carry the watermark at all.
- Text is a gray area. Google watermarks Gemini’s written replies, but the SynthID Detector and Gemini’s checker are described as tools for images, video and audio. Google has not said how families can test a block of text.
- Credentials are clues, not proof. Google itself says Content Credentials help you decide but do not definitively prove a file is AI-made.
Privacy-minded parents should also note that a Gemini check means uploading the file to Google. Skip it for private family photos you would not otherwise share.
Other ways to check if an image is AI generated
When the SynthID Detector comes back empty, a few old-fashioned habits still help. They are the same checks Google itself recommends.
- Zoom in on the details. Look for warped lettering on signs, odd hands or teeth, shadows pointing different ways and patterns that repeat a little too perfectly.
- Run a reverse image search. Finding an older original often settles the question in seconds.
- Look at the file’s metadata. An original file (not a screenshot) may show the camera model, the date and traces of editing apps.
We covered Google’s earlier on-phone approach in Android’s new tool against AI fakes, and our roundup of must-have AI software tools shows how many apps can now produce convincing pictures, which is why a checker like this is overdue.
Frequently asked questions about the SynthID Detector
Can the SynthID Detector spot images made with OpenAI tools?
Yes, if the file carries the watermark. Google names OpenAI among the partners whose AI content the portal can check, alongside NVIDIA and Kakao.
Does SynthID work on AI written text?
Not for everyday checks yet. Google marks text from the Gemini app, but its public checking tools are described for images, video and audio only.
Can I check if an image is AI generated on iPhone?
You can open synthid.com in an iPhone browser. Gemini’s Content Credentials check is on the web and Android now, and Google says it is coming soon to iOS.
Should your family bookmark the SynthID Detector?
If anyone in your house forwards viral photos, shops from social media ads or helps grandparents sort real news from fakes, yes, bookmark synthid.com today. If you rarely question what lands in your feed, the Gemini check is enough when the occasional odd clip shows up. Either way, treat the SynthID Detector as a strong clue rather than a final verdict, and keep your own eyes in the loop.
