I did a little experiment and I want others also to share their results and a possible explanation.
When you put any text in Gemini and ask, is this written by you? You will be amazed to see that Gemini is almost always highly confident (close to 100%) that the text was written by it. It is so surprising because if you repeat the process with any other prominent AI, Chatgpt, Grok, Deepseek, or Leo (these I checked), they will categorically tell you that NO, the text was not written by them, (or yes in case it is written).
Perhaps Gemini is not keeping a record of its own output as others are doing, but it also shows that the line between humans and AI is practically invisible for even data giants like Google.
I've tried it with ChatGPT and Deepseek, ChatGPT so far never admitted that the text I've submitted was written by it, it would only claim if they share similarities with AI generated texts. Deepseek however, recognized if it was written by it, I am not sure if they save such data and are able to retrace it back if asked, but for ChatGPT at least, I've tried generating a response and then asking if it wrote it and it would refuse. I am not sure if Gemini is keeping a record, have you tried generating response and asking it, just like I mentioned earlier?