AI slop is getting so good that it’s not always easy to distinguish a photorealistic forgery from a real photo. It’s no longer as easy as looking for the six-fingered renderings of just a year ago. Beyond looking for subtle clues, spotting AI images often comes down to a gut feeling that something simply looks off.
Now there’s another tool to help identify fakes, and it works with AI-edited audio and video, too. Google has opened its SynthID Detector to the public, so anyone with a Google, Apple or ChatGPT account can upload media to determine whether it’s been saved with a SynthID tag indicating manipulation using generative AI.
Google introduced SynthID in 2023 and, in late 2025, made it possible to query Google Gemini about whether images were created using Google’s AI. At the Google I/O annual developer conference this past May, the tool expanded to include detecting video and audio content in Google Chrome and Google Search.
The standalone SynthID Detector tool currently detects media generated by Google, OpenAI, Nvidia and Kakao. Not yet named is Apple, which announced it would support SynthID at its recent fall event announcing the iPhone 18 Pro and iPhone Duo.
(Disclosure: Ziff Davis, CNET’s parent company, in 2025 filed a lawsuit against OpenAI, alleging it infringed Ziff Davis copyrights in training and operating its AI systems.)
A spokesperson for Google, in an email to CNET, emphasized that working with other companies is crucial when it comes to identifying AI-generated content because no single company can solve the problem alone. They also said there’s no specific date for Apple support, but that it will be available soon.
Even so, the SynthID Detector seems already wired for Apple’s content. I uploaded a photo shot with an iPhone 18 Pro and edited using Apple’s Reframe tool, and it successfully detected the embedded SynthID tag. For an untouched iPhone-captured photo, the detector correctly did not detect a tag.
Images generated from prompts using ChatGPT were correctly flagged as including AI-generated content
To probe the level of Apple support in the tool, I created an entirely AI-generated image in Apple’s Playground app using ChatGPT’s image generation model, but the SynthID Detector did not identify it as AI-generated. It’s likely the Playground app does not yet have SynthID tagging wired in.
I also tested some photos captured using a Google Pixel 11 Pro. The company takes a two-pronged approach to image veracity on its own devices: embedding C2PA (or Coalition for Content Provenance and Authenticity) metadata in every image it creates to verify the accuracy of the original capture, and adding a SynthID when an image has been edited using AI.
Although Apple is adopting SynthID, it has not signed onto C2PA. Instead, it’s using its own Reference Image technology to store a cryptographically verified original of an image. Those shots must be captured in the Reference mode in the iPhone’s Camera app, though.
It’s also encouraging to see Google handle the media in a sensible way. When you first visit the site, you must agree to terms of service that spell out that media files you upload are processed in real time and then deleted after the results are displayed. A digital signature of the media is mapped to your user identifier and then deleted after 24 hours.
OpenAI has its own verification tool, but it only looks for OpenAI watermarks; the AI-generated samples I tested with Google’s tool came up negative on OpenAI’s site.
Any tool that helps determine genuine content is welcome, but as usual, the responsibility falls on the viewer to make that determination — if not by noticing on sight, then by making the effort of downloading the media, loading the tool and uploading it. The upside is that tracking technology such as SynthID is finally making its way into more media and paving the way for more widespread detection of AI slop.
See full bio
Discussion about this post