You rule. I was just about to deal with a work situation, and you posted this. Claude almost got me there, but it wasn't human. I used your advice and got a very usable response, with a real human touch :). Timing is everything!
I should have known Sab would come to the rescue! The interesting thing, apparently obscure to some, is that these "watermarks" and the (pending) watermark detectors are basically binary in operation. So, if you "write it yourself" and let an AI edit it, *exactly* the same way as you'd use a human editor, the detector could still say, "AI touched this." So, you're painted with a broad brush, even if you did not generate the now-fabled "AI slop" that so many seem so worried about.
Can't speak to true watermark detection, different mechanism. But the AI-writing scanner we deal with daily, Substack's own, isn't watermark-based at all, it just reads the text itself.
And it's not fully binary either. It scores how much of a piece reads as human overall, so a paragraph built on real writing can absorb a short AI-written quote or caption and still read mostly human. What doesn't survive is a piece with no real writing anchoring it, that scores as AI no matter how much it gets edited afterward.
Here's a thought... do your own writing. It will absolutely sound like you. Then use AI for editing suggestions (I never let it rewrite anything)... use what works and leave the rest. AI is so useful in so many ways but writing like me just isn't one of them. It takes less time to write it yourself than all of these gymnastics trying to make it sound human. AI is also wonderful for idea generation especially if you have it interview you.
You rule. I was just about to deal with a work situation, and you posted this. Claude almost got me there, but it wasn't human. I used your advice and got a very usable response, with a real human touch :). Timing is everything!
watermark remover link is not working says :site not found"
Trying to DM I have a question I pay for ChatGPT should I being paying for perplexity and have access to all of the ai models?
I'm here. What do you need?
After pasting and hitting Enter in Claude. What is wrong?
This is great. Thank you!
Is there a similar approach that can be applied to perplexity and Gemini?
Thanks Sabrina for all your education.
Thanks for sharing the prompts but, I am having trouble copying it on my phone.
Why are people worried about this?
I should have known Sab would come to the rescue! The interesting thing, apparently obscure to some, is that these "watermarks" and the (pending) watermark detectors are basically binary in operation. So, if you "write it yourself" and let an AI edit it, *exactly* the same way as you'd use a human editor, the detector could still say, "AI touched this." So, you're painted with a broad brush, even if you did not generate the now-fabled "AI slop" that so many seem so worried about.
Can't speak to true watermark detection, different mechanism. But the AI-writing scanner we deal with daily, Substack's own, isn't watermark-based at all, it just reads the text itself.
And it's not fully binary either. It scores how much of a piece reads as human overall, so a paragraph built on real writing can absorb a short AI-written quote or caption and still read mostly human. What doesn't survive is a piece with no real writing anchoring it, that scores as AI no matter how much it gets edited afterward.
Here's a thought... do your own writing. It will absolutely sound like you. Then use AI for editing suggestions (I never let it rewrite anything)... use what works and leave the rest. AI is so useful in so many ways but writing like me just isn't one of them. It takes less time to write it yourself than all of these gymnastics trying to make it sound human. AI is also wonderful for idea generation especially if you have it interview you.
Love love love your content and learning so much. Eventually I will need Blotato!!!! THANK YOU SABRINA. ❤️