AI can spell… but it still can’t think

By Colin Ford

One of the unexpected consequences of the AI revolution is that professional writers now find themselves having to prove that they can actually write.

For years, writers worried about blank pages, tight deadlines and difficult edits. Now there’s a new concern: “Does this sound like AI wrote it?”

It’s an odd accusation to level at someone who has spent decades honing their craft. The irony is that AI has become very good at fixing the mechanical aspects of writing. Spelling mistakes are disappearing. Grammar is improving. Sentences are becoming cleaner and more readable. That’s largely a good thing. Communication benefits when people can express themselves more clearly. But somewhere along the line we’ve started confusing good writing with AI writing.

The curse of the em-dash

Take the humble em-dash. Long before ChatGPT burst onto the scene, writers were happily sprinkling em-dashes throughout their work. Journalists used them. Novelists used them. Editors used them. They were simply another tool in the writer’s kit.

Today, however, the appearance of an em-dash can prompt suspicion.

“Ah-ha,” someone says knowingly. “AI wrote this.”

No, actually. Some of us just know what an em-dash is. (Not to be confused with an en-dash, which is the shorter one most often used as a hyphen – ironically, as it appears in the word em-dash. But I digress.)

The same thing is happening with sentence structure, vocabulary and rhythm. Writers who have spent years learning how to construct clear, concise prose are suddenly finding that their work is being judged not on its quality, but on whether it resembles the output of a machine.

The situation becomes even more absurd when you look at AI detection software. A number of these tools have flagged passages from Mary Shelley’s Frankenstein as being AI-generated. Think about that for a moment. A novel published in 1818, more than two centuries before generative AI existed, is apparently suspiciously machine-like.

Perhaps the problem isn’t the writing. Perhaps it’s the detector.

The reality is that most AI detection tools are making educated guesses. They look for patterns, predictability and certain stylistic characteristics. They don’t know who wrote the words. They don’t understand intent. They don’t have access to the creative process that produced the final draft. They’re guessing. And sometimes those guesses are spectacularly wrong!

That’s not to say AI isn’t being used. Of course it is. Most professional writers now use AI in some form. It can help generate ideas, tighten wording, check facts, identify inconsistencies and catch errors that would otherwise slip through. I use it myself.

Writing v communicating

Using AI is not the same as outsourcing thinking. The difference between competent writing and effective communication has never been grammar. It’s judgement. It’s understanding an audience. It’s recognising what matters and what doesn’t. It’s knowing which story to tell and which angle to pursue.

A few weeks ago, a client generated a press release using AI. Grammatically, it was flawless. Structurally, it ticked all the boxes. The problem was that it read more like a product manual than a news story.

The release was packed with technical detail, industry jargon and feature descriptions. It explained what the product did, how it worked and why the company was excited about it. What it didn’t explain was why anyone outside the organisation should care.

The feedback from journalists was telling. The release wasn’t badly written. It was simply too technical. There was no clear human angle, no broader context and no obvious reason for readers to engage with the story.

That’s where experience still matters.

A good communications professional knows that media coverage isn’t driven by specifications and technical language. It’s driven by relevance. The challenge isn’t explaining what a product does. It’s explaining why it matters.

AI can assemble information. Humans provide perspective. That’s the distinction that often gets lost in conversations about AI and writing.

The technology is getting better every day. It can help us write more efficiently than ever before. It can catch mistakes, improve clarity and save time. But it can’t replace experience — and it certainly can’t tell the difference between an AI-generated paragraph and an experienced writer who happens to know how to use an em-dash.

Which leaves professional writers in a curious position. For the first time in history, we’re being accused of sounding too much like the machines that were trained on the fruits of our labour.

And by the way, did I use AI to help write this? Yes. Does it make it less interesting or impactful? That’s for you to decide…