Quite frankly I disagree on almost everything. While the — or repeating the question may be legitimate red flags, the rest is not. There is a reason for that.
LLMs are trained on old material in which — was more common. Since they imitate what they know, here’s why we get all those —s. Repeating the question is a technique to reduce hallucinations, that’s why it is also quite common. Everything else is just how many people write. The average style may sound more academic and less “natural” simply because the training on academic papers usually has more weight than the training on blog posts. The rule of three is common in AI because it’s common human. Emojis were a thing in corporate messages way before AIs. The word choice highly depends on the writers’ culture, including whether they are native speakers or not. And so on.
Besides that, one can tweak the style easily. This is generated with AI using a simple prompt
AI is reshaping society—transforming how we work, communicate, and solve complex problems.
Its potential is immense, but responsible innovation is the key to lasting impact.
This is the same prompt enhanced with some tricks.
AI is seriously leveling up how we connect with audiences and streamline workflows so if you’re not tapping in you’re already behind.
Nice video, but the truth is that false positives and false negatives are so common that AI-detection techniques alone cannot be trusted without more context.
I would argue that your “enhanced” sentence is even MORE typical of AI. The first sounds like the style of a standard news article with a little corporate-speak, while the second is straight up head-of-marketing internal memo style.
Regardless, I agree that AI-detection can’t really be automated or codified. But I don’t think it matters. Bad writing is bad writing.
Quite frankly I disagree on almost everything. While the — or repeating the question may be legitimate red flags, the rest is not. There is a reason for that.
LLMs are trained on old material in which — was more common. Since they imitate what they know, here’s why we get all those —s. Repeating the question is a technique to reduce hallucinations, that’s why it is also quite common. Everything else is just how many people write. The average style may sound more academic and less “natural” simply because the training on academic papers usually has more weight than the training on blog posts. The rule of three is common in AI because it’s common human. Emojis were a thing in corporate messages way before AIs. The word choice highly depends on the writers’ culture, including whether they are native speakers or not. And so on.
Besides that, one can tweak the style easily. This is generated with AI using a simple prompt
This is the same prompt enhanced with some tricks.
Nice video, but the truth is that false positives and false negatives are so common that AI-detection techniques alone cannot be trusted without more context.
I would argue that your “enhanced” sentence is even MORE typical of AI. The first sounds like the style of a standard news article with a little corporate-speak, while the second is straight up head-of-marketing internal memo style.
Regardless, I agree that AI-detection can’t really be automated or codified. But I don’t think it matters. Bad writing is bad writing.
On macOS and iOS — is just two dashes or alt+dash
I use it often, it’s not a dead giveaway of AI.
on Android, just hold - to select a variant: – or —. On Linux, Compose+ —
I also use it quite a bit – it’s a nice alternative to commas to structure sentence flow