I May Not Like The Em Dash, But I’ll Defend To The Death Your Right To Use It

Oh, the poor em dash. Possibly the most discussed punctuation mark of the century; an unlikely martyr in the battle against the homogenisation of language and linguistic creativity in the age of artificial intelligence.

We just can’t get away from it. And yes, I’m well aware of the irony in my saying this, given the fact that I also can’t stop bringing it up. But the thing is, it’s not going away, because the problem it represents is not only not being solved; it seems to be getting worse. Much like a candle dripping wax, the issue is compounding – what was once a few drips here and there is beginning to accumulate, forming a thick layer of wax that becomes increasingly difficult to peel away from the candle stick.

The em dash is merely the symptom of the real issue at hand. That is, a fear of AI-generated writing and, now more than ever, a fear of being accused of creating AI-generated writing and passing it off as your own.

We’re Prioritising “Looking Human” Over Writing Well

When we were first introduced to ChatGPT and other early AI models, we were blown away by how well these models could write; how they could come up with responses to basic prompts and mimic human speech. And even though it still had a long way to go at that point (it wasn’t nearly as good as it is now), this sense of wonder quite quickly shifted to a rather poignant and collective “uh oh” moment.

As AI writing became increasingly difficult to distinguish from content actually written by humans, our desire to know the difference grew too. And when we realised we couldn’t always tell the difference (and that this was only going to get worse), we seemed to lose trust. Trust that content labelled as being written by humans was genuinely written by a human.

And so we began our slow descent into an abyss of mistrust and suspicion.

Nobody’s word is sacred anymore, and to make things worse, with AI detectors being so notoriously unreliable, there truly is nothing we can use to verify how (and by whom) content has been generated. Thus, the obvious conclusion and solution was to try and make our writing “look more human”.

Which is bizarre in itself. Of course, anything you or I write should look like it was written by a human, because, well, it was. And if the conclusion is that it doesn’t look like it was written by a human, the problem is very obviously not the writing itself, but the process of and metrics for evaluation.

Right?

Wrong! Well, not wrong, but for some reason that’s not the conclusion that’s been reached among our peers, and this is something we should all find quite alarming.

Instead of focusing on writing naturally and for ourselves, we’re becoming obsessed with ensuring that our writing “looks” like it’s been written by a human. And the absurdity in that is hard to explain. We’re so afraid of our writing looking like it hasn’t been written by a human that we’re removing the things that make it natural (and “human”, in many ways”).

In the pursuit of (supposedly) maintaining the quality of writing in the face of the threat of AI, we are actively decreasing the quality of our writing.

And one of the main reasons this is becoming such a massive problem is that its being exacerbated by the influence of outside judgement; the threat of the accusation.

Introducing Public Policing: Fear and Suspicion Run Rampant

It almost feels like a witch hunt. We know there’s so much “AI slop” out there that we’re constantly on the lookout. We’re suspicious of everybody, and in many respects, I think that we’ve even forgotten what authentic human writing actually used to look like without the air of suspicion that clouds everything we do.

And I don’t think anybody’s immune to it, regardless of whether or not you’re aware of the problem. I, for one, am well aware that we’re on a never-ending witch hunt (in which we don’t actually know what the witches look like), yet I can’t help but read everything with at least a hint of suspicion. Or at least an awareness that it may not be written by the attributed author, or any author at all.

This has been going on for quite a while now (and I’ve been complaining about it for some time), but it seems to be reaching new heights recently. I’ve noticed this specifically on LinkedIn (along with the rest of the professional world), but perhaps more so on public relations and journalistic platforms. That is, platforms that help connect journalists with public relations agents and experts in order to share expert commentary and so forth. The issue I’ve noticed here, however, is not simply that a lot of the writing and pitches seem to be AI-generated, but rather that there are a hell of a lot of accusations.

On LinkedIn, this was in the form of people simply posting about how everything seems to be written by AI these days. To the point where my own LinkedIn feed was basically alternating posts between terrible AI content and real people complaining about AI content.

On the other platforms, though, what I noticed was that they started presenting with AI flags, and they did this long before LinkedIn did. Ultimately, users could report something as “likely AI-generated” if it looked suspicious. And on the other side, some responses or requests for comment may have a little warning flag, alerting users to the fact that the individual in question has been reported in the past. Now, I thought the latter was an interesting approach, because it seemed to be a solution that allowed the platform to say, “hey, we’re responding to the problem” without actually making any definitive judgements or accusations. They’re not saying, “this is AI-generated”; they’re saying “some people think it’s AI-generated”. It’s pretty much the epitome of the hands up, “I donno!” emoji.

They’re giving you the option to police and report, but what about verification? Are they looking into the users who are being flagged and making human judgements before branding them with the scarlet “AI” letter? Or are they totally rejecting any sense of responsibility?

The thing is, as much as I don’t like this approach and think it’s totally counter-productive, I do very much understand why they’ve gone in this direction. It’s a tightrope to walk, and in all honesty, there isn’t really any right answer (at least, in the eyes of the public).

Do nothing because nobody can genuinely, definitively detect AI (whether by AI detection models or human oversight) – you’re not doing enough.

Mark things as definitively AI when you’re “pretty sure” – there’s no way to prove anything, so you’re making a baseless accusation and you will absolutely get in trouble for this.

Allow people to provide their opinion and share with others without deeming it true or false – better, perhaps, but a cop out, no doubt.

The problem is that we’ve found ourselves in an era of over-policing and panicked reporting. Like a wild witch hunt for our friends and colleagues, but with no more evidence than “I think” or “I’m pretty sure”.

And as a result of this overarching fear of being accused of using AI to write for us, we’re starting to change our writing. We’re straying away from our natural style and shifting into what we think makes us sound more human. People are actively decreasing the quality of the work – some even introducing errors as a supposed indicator of humanity – because we’re so scared of the beast that we’ve created.

We’re Asking the Wrong Question About Writing

But, when did good writing become suspicious?

The problem is that we’re asking the wrong question. Instead of asking whether a piece of writing is insightful, interesting, persuasive or useful, we’re asking whether AI was involved. Instead of debating ideas, we’re scrutinising punctuation, and instead of rewarding good writing, we’re becoming suspicious of it.

Of course, there are legitimate concerns about AI-generated content, there’s absolutely no denying that. Nobody wants their feeds flooded with generic, low-effort posts churned out in seconds and published without a second thought. And as a writer myself, I’m terrified of a world in which writing for ourselves as a means to express ideas and creativity becomes a thing of the past.

But somewhere along the way, the conversation has shifted to a new realm. We stopped focusing on the quality of the writing and started focusing on the process behind it, and that’s a dangerous place to end up.

Because good writing has always shared certain characteristics. It’s clear, it’s structured and it’s polished. Yes, sometimes it uses an em dash, but sometimes it doesn’t. Much like a human would. None of those things are evidence of AI; they’re simply the tools writers use to communicate effectively, things that AI models have been trained with.

The irony is that, in our attempt to preserve authentic human writing, we’re encouraging people to write less naturally. We’re removing quirks, second-guessing stylistic choices and abandoning perfectly valid techniques because somebody on LinkedIn might decide they look suspicious (or report a post as “AI slop”, but let’s not even go there).

I don’t like the em dash. In fact, it’s my least favourite punctuation mark, and I’m proud to say that my bizarre disdain existed long before ChatGPT entered the equation. But, even with this in mind, I implore you to continue to implement it in your work.

Indeed, I may not like the em dash, but I’ll defend to the death your right to use it.

Because, for goodness sake, if a writer likes em dashes, let them use em dashes! If they like semicolons, let them use semicolons. If they write in long, carefully structured paragraphs, good for them! And if they enjoy using short, punchy sentences for emphasis, right on. The goal should never be to make our writing look less intelligent, less polished or less distinctive simply to prove that a human was behind it.

After all, the question was never supposed to be “was this written by AI?” The question was meant to be, “is it any good?”