8 posts were split to a new topic: Pia Zadora thread, 80s would be starlet
I’m going to register a (doubtless futile) protest against this presumed AI tell. I see it trotted out frequently at, for example, Reddit. As a judicious user of that punctuation, I get annoyed at my humanity being questioned and dismissed by posters determined to root out AI on such a feeble presumption. Do these crusaders not recognize that AI was trained on legitimately human-written prose which includes the very punctuation they decry?
/rant
That would require them to think logically, an act which I have seen no evidence to believe is within their capacity.
[sigh] True dat.
Yeah, I know, I know. I’ve been happily using the em dash for many years, because I can type it fast (Alt+0151), and I don’t want my writing to come under a cloud—but I think within the use of the em dash by AI there are subtle tells, as I hinted above.
Yeah, this really pisses me off. I use dashes on occasion in casual writing, and now I have to stop and think about what to replace them with so that people who mindlessly believe that dashes means you’re an AI won’t be triggered.
It’s about as dumb as saying using ellipses makes you a bad writer…
Yeah ..not buying it’s a legit tell, if some one is using AI to post.
I’m personally gonna trust most people are honest and telling their truth.
If that is pollyanna-ish, I can live with it.
I see what you did there…
I also use them, and have no intention of stopping.
So an em dash is a tell except when it is not. What are these subtle tells?
There’s no single “gotcha” that proves AI use, but there are patterns that tend to show up—especially when someone relies heavily on tools like ChatGPT or Grammarly without much editing.
One subtle tell is a kind of over-smoothness. The writing is grammatically perfect, evenly paced, and polished—but a bit generic. It might feel like it’s saying the “right” things without ever getting specific or personal. Human writing usually has some unevenness—quirky phrasing, minor inconsistencies, or strong opinions.
Another sign is vague authority. AI often writes in a confident tone even when it’s being general. You’ll see phrases like “it is important to note” or “this highlights the significance of…” without concrete examples or original insight to back them up.
Repetition is also common. Not obvious copy-paste repetition, but rephrasing the same idea in slightly different ways across paragraphs. It can feel like the text is circling a point rather than advancing it.
Watch for oddly balanced structure. AI loves symmetry—three neat points, each explained in similar length and tone. Real people are usually less tidy; they might linger on one idea and barely touch another.
There’s also the lack of lived detail. If the topic should naturally include personal experience, sensory detail, or specific anecdotes—but doesn’t—that’s a hint. AI tends to stay at a “safe distance” unless prompted otherwise.
Sometimes the vocabulary gives it away. You might see slightly formal or “textbook-ish” word choices that don’t match the person’s usual voice, or sudden shifts in tone within the same piece.
That said, none of these prove anything on their own. A careful human writer can sound “AI-like,” and a heavily edited AI draft can sound very human. The more reliable signal is a cluster of these traits combined with a mismatch between the writing and what you know about the person’s normal style.
If you want, you can share a sample and I’ll walk through how it reads and what stands out.
Curiously, that response, which was of course generated by ChatGPT, doesn’t mention em dashes—and yet it uses them so frequently. I’d say it’s more an example of what AI looks like than an accurate description or what AI looks like. Seeing lots of “quotes” and italics too.
I’m kindly amazed there is no…
- Bulletized list
…because that also seems common with AI.
ETA: Based on that, one thing I think we can safely say CahtGPT is not is self-aware. Because it seemingly has no clue what it’s own tells are (the “it’s” is how you know a human wrote this, btw).
ETA2: Also, no sense of irony.
It was an impression I got reading that now-vanished thread. I don’t remember what it was exactly, sorry. It just seemed to be an em dash in a spot where people normally wouldn’t use one, even people that use it.
Serious question - why would anyone use an em-dash for internet writing when the minus key is right there?
Even if you have memorized the arcane spell to summon them with 4 keystrokes, functionally they’re no different than the minus dash, today.
AIs sure, there’s no difference in effort for them. And in some formal writing contexts, sure, I see the value. But internet posting?
Because the minus key is an n-dash. Two minus keys is an m-dash normally.
But here on the SDMB, three minus keys forms an m-dash—see?
Another frequent em dash user checking in. While I think @Johanna’s conclusion that the posts in question were copied directly from ChatGPT was astute and very likely correct, and ChatGPT does tend to make inordinate use of em dashes, that’s probably the weakest clue since human writers tend to use them a lot, too. The stylistic indicators mentioned are probably more reliable “tells”.
Ironically, I use em dashes not for formality but for the opposite reason. I often use them in contexts where, strictly speaking, a semi-colon or colon would be more appropriate. I’ll sometimes use those, too, but in general that seems too formal for a message board. An em dash creates a less formal break that is stronger than a comma.
How much harder is it to type two consecutive minuses than to type just one? Discourse automatically turns two minus signs into an em dash. I refuse to use a simple hyphen as a dash because it’s just as easy in Discourse (or most word processors) to type two consecutive hyphens which automatically turn into em dashes.
Composed the above post before I saw this. You’re quite right and apparently what we’ve all been referring to as an em dash is really an n dash. But no, the minus sign is not an n dash, it’s a hyphen. So on Discourse:
- single “-” is a hyphen or minus sign (-)
- double “-” is an n dash (–)
- triple “-” is an m dash (—)
ETA: Correction! I see that ChatGPT really is using an em dash. What I tend to use – two consecutive hyphens with a space on either side – is apparently interpreted by Discourse as an n dash.
Oh, you sweet summer child. I can tell you’ve never taken a legal writing course before. Needless to say (and as @wolfpup describes above more generally) there is a world of difference between a minus sign (hyphen) and an n-dash (which much specifically be used when citing to a range of pages).
Anyway, most word processors will automatically make an em-dash out of a double hyphen. Which can be frustrating when trying to submit a manuscript to a publisher that specifies that em-dashes should be written as two distinct dashes.
But why bother? Is anyone going to think, “That little line over there is slightly shorter than it should be - I have no idea what they’re trying to say”?
Umm … I’m a lawyer, and have studied legal research and writing. I’ve done both, including in cases that have made it before the Supreme Court of Canada. I had a paper published in a Law Review. I was a professional writer before going to law school. As I’m fond of saying, “I’ve written more books than Stephen King, but they don’t make mine into movies.” Probably because the XYZ User’s Guide wouldn’t make a good movie.
Your claim that I have never taken a legal writing class, is just plain bull.
It actually makes you look worse, then. A wizard should know better.
Dahell? Please explain.