I tried a lot of these tools, including Turnitin, and I think they are all wrong. Not because they are a bad implementation, but just because the problem is naturally impossible in a lot of cases.
There are people whose style is closer to AI, that doesn't mean they used AI. And sometimes AI outputs text that look like a human would write.
There is also the mix: if I write two pages and I used two sentences by AI (because I was tired and I couldn't find the right sentence), I may be flagged for using AI. Even worse, if I ask AI for advice and then I rewrite it myself, what would be the output? I can make a reasoning that both (AI written and not AI written) would be wrong.
> There is also the mix: if I write two pages and I used two sentences by AI (because I was tired and I couldn't find the right sentence), I may be flagged for using AI.
None of these tools are binary. They give a percentage score, a confidence score, or both.
If you include one ai sentence in a 100 sentence essay, your essay will be flagged as 1% AI and nobody will bat an eye.
It's not, but the fact that one sentence deserves a high score doesn't automatically mean that entire thing will flag false positive. Unless it's like, two sentences in total.
There are people whose style is closer to AI, that doesn't mean they used AI. And sometimes AI outputs text that look like a human would write.
There is also the mix: if I write two pages and I used two sentences by AI (because I was tired and I couldn't find the right sentence), I may be flagged for using AI. Even worse, if I ask AI for advice and then I rewrite it myself, what would be the output? I can make a reasoning that both (AI written and not AI written) would be wrong.