We’re all talking about it, we all know it – LinkedIn, and many other platforms and communications, are becoming flooded with AI-generated writing (specifically, using large language models like Chat GPT, Claude, Gemini and Copilot).
As a writer, comms practitioner and developmental editor, I’ve spent many years’ writing on behalf of organisations and individuals. I listen carefully to how people speak, the rhythm, the language, the words they choose and those they don’t. Sometimes, I’m lucky enough to be guided by a brand book, tone of voice guidelines or style guide. Most of the time I’m not, and clients are hardly aware of the work I do to understand and communicate their authentic voice.
It’s something we can all do for ourselves – and we’ll increasingly need to, to help us stand out among the AI-generated content that’s filling our lives.
In fact sometimes, writing a little clunkily can even make your writing more appealing to people because it shows that it’s a human writing. Sometimes we can’t find the right word so a ‘not-so-perfect’ one has to do. Sometimes we’re not perfectly grammatically correct. Sometimes we change our minds. And this article is written in this spirit – I’m writing it fast, without the help of AI, and it will be messy and you will pick up things that could be expressed better. But at least it will sound like me.
What’s wrong (and right) with AI?
I may divide the audience here, and I’ll be interested to hear your views in the comments.
AI can be brilliant for breaking through writer’s block, generating ideas, getting something down on the page, and for being a critical friend and thinking partner. It can help you think differently about a topic, and pull jumbled notes and thoughts into a coherent structure.
But when we use AI-generated content without editing it thoroughly (or ideally, rewriting it completely), we lose the things that make our writing worth reading in the first place: our authentic tone of voice, views and opinions.
In communications, we ideally want to encourage the reader to know, like and eventually trust us. Yet often with AI content, the reader will know something’s off – even if they’re not consciously aware of this – and it will affect your credibility. The writing doesn’t sound like they think you sound, or it feels too similar to other people’s.
I predict that if this continues, people will lose their audiences completely. Many people have already lost me. I immediately scroll past posts and articles that show telltale signs of being AI-generated. It’s almost as though my brain switches off. Strangely, and I’m sure there’ll be research about this soon, it feels like cognitive overload, even though AI’s sentences tend to be short, separated by line breaks, and supposedly easily skimmable. Those are all things that, as writers and comms practitioners, we advise clients to do – but not to that extent!
The way you speak and write, your particular way of seeing things, is unique. If you hand that over to AI completely, you’re losing trust and connection – and for business readers, also your competitive advantage.
How to spot AI-generated writing – and what to avoid
I’m still learning and observing this, and I’d love to hear your thoughts, but here are some of the things I’ve spotted:
1. Faux-conversational bridges and filler phrases: “Here’s the thing…” ”But here’s the truth…” ”I’ve been there…” ”Don’t get me wrong…” ”At the end of the day…” AI’s attempt to sound authoritative yet relatable. We all use some of these occasionally, but they’re everywhere!
2. The paragraph thing: I can only show this, not tell, so here goes –
Then there’s the paragraph thing. You’ll see it everywhere now.
One short sentence.
Then a paragraph break.
Another short sentence.
Another break.
Perhaps a longer sentence here where someone really wants to nail the main point, but the sentences are never very long.
Few paragraphs are longer than two lines.
And in many cases they’re really, really long posts that could be an article.
It’s meant to create impact and rhythm, but it ends up feeling mechanical. Real human writing has a messier, more organic flow. We vary our paragraph lengths because we’re following the natural shape of our thoughts, not a formula.
3. Faux authority 1: flat statements of truth without any softeners or qualifications like, ‘in my experience’. Eg Clarity isn’t optional. It’s foundational.
4. Faux authority 2: overuse of statements where you balance one thing against another. Symmetrical statements perhaps. Eg Strategy without execution is just planning. Execution without strategy is just activity.
5. Titlecase For All Headings Meaning Every Word Starts With An Uppercase: this is something that divides people in the real world. I prefer sentence case or first letter uppercase where only the first letter and proper nouns are uppercase. Whichever you prefer, titlecase is definitely a sign of AI.
6. Lines between sections: this happens with longer content like reports. It’s not that this is a bad thing, it can make it clearer, but it’s definitely an AI characteristic.
Many of these devices are ones that I use and recommend, and can be helpful in the right place. But in many articles they’re all piled in there like mixed waste in a landfill site.
What I’ve learned from writing and editing with people and with AI
If you’re using AI for writing and not editing it, I know you aren’t trying to deceive anyone. You have something to communicate, and AI helps you make that happen. You may be used to having great insights, advice and questions, but not having the time to write them down.
So here are a few tips that may help you to edit your writing, before you hit publish:
1. Read what AI has written out loud: not in your head – actually speak the words. If you stumble, if phrases feel awkward, if you’d never say it this way, then rewrite it.
2. Hunt for those filler phrases (see above): to make sure you don’t miss them, you could even use the search function. Unless these are genuinely part of your speaking voice, delete them. Think about how you’d introduce that thought if you were talking to someone.
3. Vary your rhythm naturally: instead of single-sentence paragraphs scattered throughout, let your paragraphs breathe in different ways. Some thoughts need three sentences, some need far more. Follow the natural shape of your idea.
4. Let your imperfection show: real human writing has a bit of roughness to it. We use dashes more than we should, we start sentences with “And” or “But” (which is grammatically OK by the way – and don’t let anyone tell you otherwise), we let our personality leak through, and our sentences are sometimes longer than they should be.
5. Use your actual vocabulary: if you’d never say “leverage” or “ecosystem” or “innovative solutions” in real life, don’t write them. Use the words you actually use. Your writing should sound like a slightly more thoughtful version of how you speak, not like a different person entirely.
6. A practical exercise: take an AI-generated piece of copy and rewrite it as if you’re explaining it to a friend. You could even record yourself. What words would you use? What would you emphasise? Where would you pause? Compare the two versions. You’ll see what needs to change.
7. Prompt AI well: this is the real skill in using AI, but one for another blog. Don’t just accept what it provides you. Give it more feedback. Question it. Give it examples of what you do and don’t like. Upload your tone of voice guide or brand book – or if you don’t have one, create one no matter how rough.
And this is the paragraph where I’m meant to finish with a startlingly insightful and authoritative observation, but I’m not going to. Yet another way you can tell that this is written in a hurry by a human. But I hope this has been helpful and I’d be interested to hear about your experiences of AI.
What are the tell-tale signs for you? What are your pet peeves and what do you think it does well? What have been the benefits for you of using AI?
Photo by Feodor Chistyakov on Unsplash, edited.