OpenAI appears to be preparing to introduce watermarks on AI-generated images in ChatGPT, based on findings from an APK teardown of the app’s latest beta version (v1.2025.196). The feature, currently in development, may apply watermarks to all generated images, particularly for users on the free tier. However, indications suggest that paid users could be given the option to save images without the watermark.
The APK teardown — a process where in-development app code is examined for unreleased features — revealed a string labeled “image-gen-watermark-for-free,” which implies that watermarking may be limited to free accounts. This aligns with an earlier leak from April 2025, when a watermark feature labeled “ImageGen” was first spotted by a user on X (formerly Twitter).
Screenshots from the activated feature shared by AndroidAuthority show the standard image interaction options — Edit, Select, Save, and Share — along with a new option in the three-dot menu: “Save without watermark.” While this feature was momentarily enabled even for a free-tier account during testing, it’s likely that access will be restricted to subscribers once the feature is officially launched.
Image by AndroidAuthority
There is still some uncertainty about how the watermarking system will be implemented. One possibility is that all users will see watermarked images by default, with only paid users able to remove the watermark. Another scenario could involve allowing all users to save unwatermarked images, but through a more complicated process that discourages widespread casual use. This strategy could help OpenAI balance transparency and branding with user flexibility.
It’s important to note that this watermarking feature has not been officially announced by OpenAI, and its inclusion in the final release is not guaranteed. APK teardowns offer insights into features that developers are experimenting with, but they do not confirm final implementation.
OpenAI continues to experiment with a range of new features for ChatGPT, including Study Together modes, Image Styles, and subscription changes like a potential annual plan. The watermarking feature adds to this list of in-progress enhancements, potentially signaling a broader effort to manage the traceability and use of AI-generated content.
Until OpenAI makes an official statement, these findings remain speculative. Further updates are expected as development progresses.
Filed in . Read more about AI (Artificial Intelligence), ChatGPT and OpenAI.