SynthID
Fewer than 5 people recommend it so far, so there's no written verdict. Here is what commenters said.
4recommend
10warn against
Nov 20, 2025first mention
Oct 7, 2026latest mention
Why people recommend it
“SynthID is a better standard for designating something was created with GenAI”
Show 2 more
“It is, because of SynthID.”
Caveats and dissent
“It's a real shame Google isn't more transparent about how it actually works”
Show 2 more
“It absolutely does. It constrains high-entropy word choice so it can "store" other things in your text.”
“Of course that's impractical because a million people could quickly receive the image”
Often compared with
Threads where it came up
- OpenAI Adopts Google's SynthID Watermark for AI Images with Verification Tool 6 mentions
- SynthID Detector 2 mentions
- Spymarks, not Watermarks 2 mentions
- Anthropic's 'Watermark' Text Adulteration in Claude Is a Perversion of Writing 2 mentions
- Discovering, detecting, and surgically removing Google's AI watermark 2 mentions
- SynthID 2 mentions
Data: synthid.json. Quotes are verbatim; the verdict and theme labels are written by a model from these comments, and every clause links to its sources.
All comments 28
“Deepwalker cracked this one long time ago”
“It's a real shame Google isn't more transparent about how it actually works”
“SynthID is a better standard for designating something was created with GenAI”
“It absolutely does. It constrains high-entropy word choice so it can "store" other things in your text.”
“Tech like SynthID I see a net positive”
“actually reading the the synthid paper showed me that my understanding was wrong”
“SynthID should not distort the output of an LLM any more than the default temperature settings do already.”
“it feels like it will be difficult to remove watermarking on bigger text”
“Of course that's impractical because a million people could quickly receive the image”
“SynthID has been totally busted open”
“It’s definitely hackable”
“Seems inferior to C2PA”
“SynthID payloads work the same way”
“There is an approach that apparently works and is based on spectral analysis of the images.”
“https://synthidbypass.com/”
“and the previous SynthID: https://deepmind.google/blog/identifying-ai-generated-images-with-synthid/”
“a meaningless signal like this deserves such a demise”
“It's not really hard as it looks”
https://deepmind.google/models/synthid/
“SynthID is removed from images by passing the image through a diffusion model”
“Looks like there's a lot more info here, at least about the text version.”
“Google watermarks their images”
“google and maybe others put a watermark in each image”
“SynthId is used to watermark text as well, and code is open-source”
“It is, because of SynthID.”
“specific watermarking in the images themselves”
This is what C2PA is trying to do: https://c2pa.org/
“It's called Synth ID. It's a watermark that proves an image was generated by AI.”