In all the chatter¹ about dashes revealing the use of generative AI such as ChatGPT, I’ve seen two things missed.
First, confident assertions about em-dashes — like these — being ‘correct’ don’t make any allowance for regional variations. Em-dashes remain common in formal US styles guides, but shorter en-dashes – with spaces either side – are most usual in the UK, for example.
And it’s those pesky spaces that are the other niggle. I keep seeing people declare with absolute conviction that em-dashes shouldn’t have spaces either side, but while that is becoming common in practice — due to knowledge gaps, technical limitations and laziness — the established convention had been to use hairline spaces.
× With em-dashes — like these — full spaces are wrong²
× With em-dashes—like these—no spaces is also wrong² (and ugly)
✓ With em-dashes — like these — hairline spaces are right³
How do you type hairline spaces? That’s the problem: you can’t, easily. I have a macro set up which syncs to all my devices so that when I type ‘//em’, I get a hairline space (Unicode glyph U+200A), then an em-dash, then another hairline space.
And you know damned well AI didn’t write this because even it can’t be this boring.
¹ In short, there was an assertion that em-dashes in copy revealed that you used gen-AI to ‘write’ it, and the reason that could be true is that they’re trained on a corpus that includes the use of em-dashes, so it regurgitates that… even though normal people tend not to use them.
² For a given value of ‘wrong’; I take a very non-prescriptivist approach to language, but I find convention and the tension around it interesting.
³ For a given value of ‘right’. There are many different kinds of space characters, and different style conventions would mandate different ones. You could start here if you’d like.