The Soul Of The New Machine Is Not Mine

One of my more cherished possessions is a writing sample from when I was in Grade 5, where I used a fountain pen for a composition. It was an assignment centred around the theme of “Childhood.” This little piece marks the beginning of my writing journey, which has since grown to include 26 books, hundreds of articles, and thousands of tutorials.

The first paragraph, to me, is important:

“I was born in Vancouver General Hospital and I was 6lbs,50 oz. It was a cold November day of 1952.”

At the age of 10, in 1962, there was no way I could have checked Google or asked ChatGPT to confirm that it was a cold day. A quick check of the monthly average for November 1952 reveals that the temperature was 9.2 °C (49°F), and it may have been raining.  This was the first, faint, glimmering of my finding my “voice” and telling a story. 

I am coming across more and more postings and articles where the respondents claim AI wrote it. In many cases, this is absolutely true. AI can “write” but, in many respects, it is the voice of a new machine, not the author. If any of you have read my books or articles, attended one of my presentations or taken one of my LinkedIn Learning courses, you don’t need to see my byline. You just know, “Yep, that’s Tom.” I can say the same for Ezra Klein of the New York Times. Read his thought pieces and then listen to his podcast. You know it is Ezra. The same goes for Kara Swisher, whether it is her book- Burn Book- or her podcasts- On with Kara Swisher or Pivot with Scott Galloway- you know it is Kara. They sound the same in print and broadcast.

Here’s another example. I have a close friend whom I have known for well over 25 years. When I read his work. I instantly know that it's him. It sounds like him because, as we read, we sound out the words in our mind. If he were to have an AI write his stuff, I would instantly know that it is not him.

So what is this voice I am talking about? It is the author’s attitude toward the subject matter and is influenced by tone, diction and syntax. Tone is that attitude which, in my case, is somewhat relaxed, direct and authoritative. Diction is word choice used to shape the entire feel of the piece, and syntax is the arrangement of words into phrases and sentences. They all combine into that voice that is me, and it doesn’t come naturally.

Back in 2003, when I was writing my first book for New Riders, my editor kept bouncing back the first chapter, telling me to rewrite it. After three bounce-backs, I emailed her and told her to back off. Her reply was blunt. She was going to decide within a month whether to kill the contract. The issue was that I hadn’t found my voice. My writing style was flippant with a heavy dose of “Aren’t I clever?” She concluded by offering to work with me to find it and warned me she was going to be vicious. I still bear the scars from that intense month, but it set the course for my career.

The “voice” that emerged was mine: an experienced, pragmatic, and sometimes blunt educator. That editor showed me I could be blunt in my way, such as this line from a recent piece, AI’s Aversion To Risk, I wrote for the UX Collective: “Designers becoming 'curators' is very dangerous thinking.”. She showed me how to rephrase and tighten up sentences and how to inject my personality into those sentences using a “show, tell, show” structure.

To demonstrate AI‘s inability to match your voice, I fed an article I had written- AI’s Aversion To Risk- into ChatGPT and Gemini. I asked each model to analyze my voice. Then I asked them to write the first two paragraphs of the article in my voice. The results were unsurprising.

Here are the first two paragraphs of the original article:

A couple of weeks ago, I came across a LinkedIn post by Eric Edwards that marvelled at the sheer brilliance of the Open Spaces Burberry ad. He concluded the post by writing:

"But if it came out today, a lot of people would assume it is AI. No questions asked. And that’s the unfortunate part, not that AI exists, but that our instinct to be amazed by human craft is fading. Feels like we are losing the thrill of the impossible a little bit. Oh well. Price of progress?"

I couldn’t agree more with that observation and have to ask: “Are we, as UX Designers and creatives, willing to pay that price?”

Here is ChatGPT’s attempt:

If you’ve played with any of the new AI-powered design tools lately, you’ve probably noticed a peculiar pattern: everything looks... safe. Not bad, not broken—just comfortably average. Rounded corners, muted colours, perfectly aligned grids. The layouts are competent, even polished, but they rarely surprise you. They’re designed to appeal to the broadest possible audience, which is exactly the problem. In its current form, AI isn’t a creative partner—it’s a risk manager.

What we’re seeing is less a reflection of design intelligence and more a mirror of historical consensus. These models are trained on oceans of existing work, most of it optimized for scalability and approval, not experimentation. As a result, the outputs tend to regress toward the mean. Innovation, by its nature, lives at the edges—on the margins of taste, comfort, and even failure. But AI doesn’t like margins. It prefers certainty, predictability, and a data trail. In short, it avoids risk like a UX designer avoids a carousel.

… And here is Gemini:

Spend five minutes with any of these shiny new AI design tools and you’ll start to notice something: everything they spit out feels... familiar. Not wrong. Not broken. Just safe. Rounded corners. Symmetrical grids. A decent enough font pairing. It’s as if the entire design brief was “Don’t scare anyone.” The results are technically fine, but creatively flat—like they’ve been run through a corporate brand manual a dozen times before they ever saw daylight.

And that makes sense when you realize what’s going on under the hood. These tools aren’t inventing anything—they’re remixing history. They’re trained on oceans of existing designs, most of them already watered down for maximum approval. So, instead of pushing boundaries, they steer hard toward the middle. The problem? Great design doesn’t live in the middle. It lives out on the edge, where things might break, fail, or—god forbid—make someone uncomfortable. But AI doesn’t do uncomfortable. It does average.

I am also a power user of Grammarly. It, too, has also swung over to AI for writing, which, to me, moves the application solidly away from its “knitting”. Grammarly is phenomenal for catching grammatical and spelling errors. It also offers you the opportunity to “rewrite” paragraphs in a variety of styles ranging from formal to friendly. What I appreciate about this feature is that I can either accept the change and subsequently edit it or heave it into the trash. As an experiment, I told Grammarly to write 2 paragraphs on the topic of AI’s Aversion to Risk. Here’s the result:

Artificial intelligence, for all its transformative power, is often far more risk-averse than the humans who create it. While AI systems are lauded for their ability to process data at lightning speed and make decisions based on vast amounts of information, they are typically programmed with a strong preference for caution. This aversion to risk is not accidental; it is hardwired into their algorithms by designers who must account for safety, reliability, and the unpredictable consequences of machine-led choices.

This risk-averse nature manifests in many ways, from self-driving cars that hesitate at complex intersections to financial trading bots that pull back at the first sign of volatility. Unlike humans, who can weigh emotional, ethical, and contextual factors in their decision-making, AI evaluates risk based on programmed rules, historical data, and statistical probabilities. The result is often a conservative approach, where minimizing potential harm outweighs the pursuit of bold or unconventional strategies.

Not one of those is me. They don’t reflect my writing style. The text produced is also risk-averse. Both ChatGPT and Gemini had the chance to study my “voice,” and neither picked up on “Designers becoming ‘curators’ is dangerous thinking.” And the first-person viewpoint is totally absent.

Conclusion

Letting the various AI models do your writing is “Dangerous Thinking”. Copying the output, slapping your name on it and then uploading it somewhere, to me, is nothing more than “Digital Plagiarism”.

Where they shine is the ability to compile data and facts into an understandable form. For example, feed marketing data into ChatGPT and Gemini and then ask it to provide an Executive Summary or Analysis of the data. The result is more formal and can be circulated to stakeholders because it aligns with their expectations. Another example would be NotebookLM from Google. This AI allows you to feed a variety of documents, ranging from Google Docs to YouTube videos, into NotebookLM. From there, you can ask it to provide you with a number of documents ranging from Reports to an Audio Overview. On top of that, there is a chat feature where you can interact with NotebookLm to gain a deeper insight into the sources. Best of all, NotebookLM is restricted to only the source material, and when you quit any “training,” it is deleted.

AI models can mimic thinking, but they’re not quite like you. They’re handy when you need to add some variety to a monotone voice. However, this “voice” of the new machine doesn’t have the special human touch that makes you, well, you… and no one else.