Not long after I started my first writing business, I landed a plum assignment with a big consulting firm to develop white papers for them. They had invested in quality research, conducted multiple interviews with people in the target audience of high net-worth individuals, and had a panel of internal subject matter experts to interview. The hard part was done.

They needed me to extract the necessary information and mould it into a persuasive document that didn’t ‘sell’. The goal was to build trust with their audience and convince the reader they were the only reasonable choice for assisting with personal financial challenges of $1 million or more – in 12 pages or less.

I delivered the first draft and got immediate approval. The only caveat was the white paper had to be approved by the CEO. The project manager assured me it was a formality.

He couldn’t have been more wrong.

It turned out the CEO fancied himself something of a wordsmith. He edited my copy, adding four pages, strong brand messaging, and hard calls to action.  Worse, he cluttered the report with hyperbole, jargon, unintentional humour, and more gumpf than I thought possible.

I won’t go into all my trauma the drama, but the project manager resigned after the fourth draft. A new project manager encouraged me to let the CEO have his cliches and misplaced metaphors but to ensure it would still appeal to those high net-worth individuals they were desperate to attract. On the sixth round of edits, the project manager made the preposterous request to “make it sound like you wrote it in the first place”. I suggested we go back to the original draft.

They did. They never called again; I wasn’t sad.

The problem with replacing your writers with AI

When I see AI-generated writing, I get flashbacks to those first rounds of editing with the CEO. It’s full of confidence – smoothly parsing out corporate speak but in a very deliberate structure that’s easily replicated. You can almost anticipate what’s coming next and I promise you it won’t set anyone’s world on fire. There’s no colour, no sharp edges, no nuance, memorable stories, or personal insight that makes it worth reading.

Despite this, writers all over the world are being challenged, or even worse – accused of using AI to do their job for them.

Anyone with a keyboard and an opinion can weigh in, which pretty much means everyone. A lot of uninformed ‘expert’ opinions are piling on with more certainty than Steve Jobs on iPhone launch day.

Ironically, producing more of the ‘same, same, same’ content isn’t going to impress the AI agents.

It turns out humans are bad at detecting AI-generated content

As I wrote in my post about em dashes, detecting AI-generated text isn’t as easy as people think. As the AI agents improve – and learn what people don’t like – it’s getting even harder.

Consider this: even with AI detection software, humans are only slightly better than chance at knowing what was written by a person and what was generated by a machine. One study published at the National Library of Medicine, showed the overall accuracy of human detection of AI text was 19%. In other words, identifying AI text is a thumb suck, little more than a wild guess.

What content creators and writers are dealing with is nothing more than opinion dressed up as outrage.

How to detect AI-generated writing, according to an expert

I asked a subject matter expert about how to tell if something had been written by AI. My buddy Claude said common giveaways fall into four main areas:

Structure

  • Bullet points
  • Headers, even for short sections
  • Groupings of three
  • Sandwiching information: the old, “Tell them what you’re going to tell them, tell them, and tell them what you just told them.”

Language

  • Using specific words, especially “delve”, but also including “tapestry”, “nuanced”, ‘multifaceted”, “robust”, “leverage” and “underscores”.
  • Heavy use of hedging stacks in one paragraph to soften the language and make transitions, such as “It’s important to note that, “It’s worth mentioning that”, and “It’s crucial to consider.”
  • Using several words when one verb will do, such as “the facilitation of communication” instead of “communicating.”
  • Overly balanced conclusions, that don’t take a firm position, such as “ultimately, it depends on your specific needs/context/goals.”

Tone

  • No point of view
  • Enthusiasm without personality
  • No tangents or asides
  • Perfectly uniform confidence.

Content

  • Suspiciously thorough (even though Claude didn’t mention the em dash as a tell-tale sign)
  • Vague specificity with no concrete detail, (e.g., “various studies have shown…”)
  • No firsthand experience
  • Overly clean, bland examples.

Writers are their own worst enemies

If we’re so bad at differentiating what people write compared to what AI agents write, then why is the debate raging? Well, the internet never needed truth, and no amount of patient explanation is going to change the opinion of the misinformed. Once the mob latches on, it’s best to keep your head down and keep doing what you know is right.

But the bigger problem is writers are taking the bait. They’re hooking into the drama and letting the masses decide what’s good and what’s not. They’re trying to protect themselves from judgement by a public unqualified to judge.  In practice, they’re dispensing with their favourite punctuation and changing how they structure sentences and paragraphs. I’ve seen more than one writer admit to intentionally inject spelling and grammar errors into their copy as misguided proof it was developed by a real person. I’m flabbergasted.

Here’s my advice to writers, content creators and content marketers. Block the noise. Don’t feed into the hype. Work on being excellent. Hone your craft, tell great stories, and lean into your individual voice. Too many writers have dedicated far too much time trying to convince people who won’t be convinced what is right. For the love of all that’s good, please don’t intentionally inject mistakes in your writing so people won’t think it’s generated by AI. Losing your professional self-respect is not the answer.

If you want help creating high-value content to support your brand and marketing initiatives, give me a shout or get in touch with SALT. We have a crack team of human writers and the quality control processes to ensure your content is as close to perfect as humanly possible.