Everyone is asking how you even watermark plain text. There is nowhere to hide a mark in a sentence.
You do not hide anything. You bias the word choice.
When AI writes, most spots have several equally good words. Nice, lovely, great, fine. A secret key quietly nudges it toward a hidden set of them, across the whole thing.
Over enough text that leaves a fingerprint. Far more of the preferred words than chance would give you. Invisible on any single line, only visible over length.
Which is also why rewriting it in your own words wipes it out. You swap the model's biased picks for your own even ones.
@gregorykennedy Fake personalization is worse than none, it just proves that a bot wrote the email.
I think what matters is relevance, show me u get the problem not my hobby
@kimindiehacker Actually true.. but u have to be careful about what u mean by fundamentals.
It is usually the knowledge of how coding works in general and the logic behind it that matters, not the syntax or the basics of a language.
@matthias_mrc many people quits on the flat part of the curve, cuz they dont realize that marketing is backloaded which means the result lag the work by months.
In SEO there is one answer to who ranks number one. In AI search there isn't.
AI citations are probabilistic. Ask the same question twice and you can get different sources.
Part of this is a setting called temperature. It controls randomness. Low temperature gives near-identical answers, high temperature gives very different ones each run.
So you do not get "ranked" by AI the way you rank on Google. There is no fixed slot to hold.
What you get instead is a probability of showing up.
That is why the goal is AI visibility, not AI rank. You are not winning one position. You are trying to be a likely answer across many runs.
@forgebitz Like Latex replaced docs. And I think so it's a good thing.
Because then Agents can easily read and edit the docs but u still have to know what to write in it