Expert Opinion: You can always tell

James McCracken, a fractional CMO and Chair of CIM North West, explains the problems with using AI to write marketing copy, and how to spot it in your own.

I spent an afternoon last week going through a client’s website removing a single word.

Vendor.

It was everywhere. And no British marketer says vendor. We say supplier. It is a small thing, but it was a tell. The copy had been written, or heavily helped, by AI. And once you notice one tell, you notice all of them.

The dashes, for a start – long ones, everywhere, where a comma would do. The word “journey” bolted onto everything: the buyer journey, the customer journey, the transformation journey. “In today’s fast-paced digital landscape.” “Leverage.” “Robust, seamless solutions.” Three things in every list, always three, because the model likes the rhythm. Copy that is grammatically perfect and completely forgettable.

I am not against AI writing. I use it most days. It is a good first draft, a good sparring partner, a good way to get unstuck on a Monday morning. The problem is not that AI writing is bad.

The problem is that it is the same.

When everyone uses the same tool, everyone sounds the same.

Feed the same prompt into the same model and you get the same voice out. So when every firm in your sector does it, you all end up sounding identical. Interchangeable. And marketing that sounds like everyone else cannot do the one job marketing exists to do, which is to make you the obvious choice. Sameness is invisible. Invisible does not win the work.

There is a second cost, and it is quieter. Trust. Buyers are getting quicker at spotting the machine, even when they could not tell you exactly what tipped them off.

Something reads a shade too smooth, a shade too generic, and it nudges them, before your actual argument has landed. In a considered purchase, where trust is most of the decision, that is expensive.

And there is a neat irony in it. As more buyers ask ChatGPT and the like who they should hire, the businesses that get named are the ones with a distinct, human point of view. Not the ones who sound like the model’s own factory setting.

What to actually do about it

Keep using AI. Just stop letting it be the last word. Before anything goes out under your name:

Hunt the tells. Vendor, leverage, robust, seamless, “in today’s…”, the long dashes, and any list that comes in threes for no reason. Cut the throat-clearing – “it is important to note”, “when it comes to”, “in conclusion”. They add nothing.

Then do the one test that catches almost everything: read it aloud.

Anything you would never actually say to a client across a table does not belong in your marketing. If it sounds like a press release being read by a robot, it is one. Rewrite it the way you would say it.

The businesses that come out of the next few years ahead will not be the ones that used AI the most. They will be the ones that still managed to sound like themselves. Your voice is one of the few advantages a competitor cannot copy with a prompt. Do not hand it over.