ChatGPT's version of "Your site's AI-written content is an avocado bathroom suite"

This article is written by ChatGPT based on a prompt from me. The purpose is to illustrate the difference in voice between a human writer and an AI writer, and it should be read in conjunction with the actual article I wrote on this topic.

For the record, I did not feed the AI the content of my article. I gave it only this prompt:

Write a 1700 or so word blog post for Nevado's website about AI writing. Say how AI writing is useful to businesses. Cover how changes in AI models lead to changes in AI writing styles which can quickly date writing produced with the last LLM. Give examples of popular AI writing tics. Caution the reader on how this could mean an investment in content over the last X months could be outdated sooner than in the days of human-only writing. Mention that there have always been trends in human writing too. Sign off by advising the reader to (a) become more aware of the role of AI writing and see if they agree with me, (b) to talk to people in their organisation to form a consensus view and (c) to understand the importance of voice in writing and how AI can work for and against this given how perceptive humans are at picking up inauthenticity. Throughout use the analogy of a 1970s bathroom suite dating rapidly because the avocado colour goes out of fashion.

The avocado bathroom problem with AI writing

AI has made it easier than ever for businesses to produce written content. But could it also be making that content go out of fashion faster than ever before?

The trouble with avocado

Imagine it's 1975. You're renovating your bathroom and you've decided to make a statement. You've looked through the brochures, visited the showrooms and settled on a beautiful new bathroom suite in avocado green.

It's fashionable, contemporary and quite possibly the envy of your neighbours. You've spent a considerable amount of money having it installed, and you're confident you've made a good investment.

Fast forward fifteen years. The bathroom is still in perfect working order. The taps work, the bath holds water and there's absolutely nothing wrong with the toilet.

Unfortunately, it's avocado green.

What once looked like the height of sophistication now looks unmistakably dated. Worse still, changing it means ripping out something that's functionally perfectly good. You didn't necessarily make a bad decision in 1975. You just invested in something whose appearance was more closely tied to a particular moment than you realised.

I think something similar is beginning to happen with AI-generated writing. And for businesses investing significant sums in content, it's worth paying attention to.

AI is genuinely useful for businesses

Before I go any further, I should say that I'm not making an argument against AI writing.

Far from it. The benefits are substantial, and many businesses are already enjoying them.

Writing has traditionally been expensive and time-consuming. Creating a set of website pages, producing a newsletter or maintaining a regularly updated blog requires people with knowledge of the subject, an ability to communicate and the time to do it properly.

AI has dramatically reduced some of those barriers.

It can turn rough notes into coherent prose, suggest structures for complicated documents, summarise technical subjects, adapt existing content for different audiences and generate first drafts in seconds. It can make a reasonably good writer considerably more productive and help people who struggle with writing communicate their ideas more clearly.

For a small business without a marketing department, these capabilities can be transformative. For larger organisations managing enormous quantities of content, the efficiencies can be equally significant.

And the quality is often impressive. Given good instructions and appropriate source material, today's leading language models can produce writing that's grammatically excellent, well organised and apparently thoughtful.

But there's a difference between writing that's good and writing that sounds good. And another difference between writing that sounds good today and writing that will still sound good in five years.

That's where the avocado bathroom comes in.

The distinctive sound of AI

If you read a lot of business content, you've probably begun to notice certain recurring patterns.

Some words seem to appear with extraordinary frequency. Businesses are forever "delving into" subjects, "navigating the complexities" of their industries, "unlocking potential" and operating in an "ever-evolving landscape".

Apparently everything is "pivotal", "transformative" or "a testament to" something.

Then there's the particular sentence construction that explains what something isn't before revealing what it really is:

"It's not just about technology. It's about people."

"We're not just building websites. We're creating experiences."

"Success isn't simply about reaching your destination. It's about the journey."

Perfectly legitimate constructions, of course. But once you've spotted the pattern, it's difficult to stop noticing it.

There are other giveaways. AI has shown a fondness for em dashes, neatly balanced lists of three, excessively enthusiastic introductions and concluding paragraphs that earnestly summarise everything you've just read.

Consider this entirely fictional piece of website copy:

"At Acme, we're passionate about empowering businesses to unlock their full potential. In today's rapidly evolving digital landscape, we deliver innovative, scalable and future-ready solutions that help organisations navigate complexity with confidence. We're not just a service provider; we're your trusted partner on the journey to success."

It's polished, professional and almost entirely devoid of meaning.

Of course, human writers produce plenty of similarly empty corporate waffle. That's partly why AI is so good at reproducing it. Language models learn from human writing, including all our clichés, conventions and bad habits.

But AI can amplify those habits to an extraordinary degree.

There's evidence this is changing the language we encounter online. An August 2026 study by Pew Research Center, analysing approximately 490,000 webpages, found that several linguistic features associated with AI writing had become considerably more common since ChatGPT's introduction. Em dashes appeared about twice as frequently, while the "it's not just X, it's Y" construction had nearly tripled in frequency, albeit from a low base.

None of these characteristics proves that a piece was written by AI. But collectively, they contribute to a style that increasing numbers of readers recognise.

And recognition is important.

When the models change, the fashion changes

Here's the aspect of AI writing that I find particularly interesting.

AI writing doesn't have a fixed style.

Different language models have different tendencies. Even successive versions of the same model can write differently, reflecting changes in training, fine-tuning and the instructions governing their behaviour.

One generation might favour elaborate, slightly grandiose language. Another might produce something more conversational and restrained. A third might be better at avoiding formulaic constructions altogether.

These differences aren't necessarily deliberate attempts to create a fashionable new voice. But they can have much the same effect.

When millions of people use the same handful of AI models to produce content, those models' stylistic preferences become unusually influential.

An AI writing habit can spread through business websites, newsletters and social media at remarkable speed. Then, as models improve and users become more discerning, those habits may begin to disappear from newly generated material.

The result is a succession of recognisable styles.

And here's the problem: the writing stays where it was published, even when the model that produced it has moved on.

That website you painstakingly filled with copy eighteen months ago might contain phrases that the latest AI models no longer favour. At the time, the language seemed sophisticated. Today it might sound unmistakably like an earlier generation of ChatGPT.

Readers may not consciously identify which model or generation produced it. They may not even realise why the writing feels familiar or slightly tired.

They just recognise the avocado bathroom.

There's an additional complication. We're becoming more sophisticated readers of AI-generated material ourselves. What seemed extraordinary when ChatGPT first appeared is now commonplace. What once impressed us with its fluency may now irritate us with its predictability.

So the perceived quality of your content can change even when not a single word of it has been altered.

But writing has always followed fashion

It's important to keep some perspective.

Human writing has never been immune to trends.

Look through advertising from the 1950s and you'll encounter a kind of confident, declarative salesmanship that can seem almost comical today. Corporate brochures from the 1990s were often full of earnest references to excellence, quality and customer satisfaction.

The dot-com era brought its own vocabulary of disruption and innovation. Later, every ambitious start-up seemed to be "on a mission" to reinvent something.

Even the fashion for brands to sound irreverent, informal and relentlessly conversational has become a recognisable period style. There's only so much personality you can inject into a message about a delayed parcel before the jokes begin to sound contrived.

Good writers have always absorbed influences from other writers. Copywriters borrow techniques from successful campaigns. Journalists adopt the rhythms of their contemporaries. Businesses imitate competitors they admire.

In that sense, AI is simply another influence on the way we communicate.

But there's a difference in scale and speed.

Human writing fashions traditionally spread through a diffuse collection of writers, editors, publications and cultural influences. AI allows a relatively small number of models to influence an enormous volume of new writing almost simultaneously.

When their output changes, the prevailing style of freshly produced content can change remarkably quickly too.

It's rather like having millions of bathrooms installed from the same catalogue, followed eighteen months later by a new catalogue that makes the previous one look old-fashioned.

The hidden cost of content that dates

This matters particularly for businesses because writing is an investment.

Imagine your organisation has spent the past twelve months rebuilding its website. You've commissioned dozens of pages, rewritten product descriptions, created case studies and published a substantial collection of articles.

Perhaps you've used AI to accelerate the process. Perhaps an agency or freelance writer has used it on your behalf.

The resulting content might be accurate, helpful and performing well in search engines. In every conventional sense, the investment has been successful.

But what if its language becomes dated much sooner than you anticipated?

Traditionally, a well-written page about a relatively stable subject could remain useful for years, requiring occasional factual updates rather than a complete rewrite.

Now, some businesses may find that perfectly serviceable content begins to sound unfashionable within a much shorter period.

I'm not suggesting that every AI-assisted article has an eighteen-month expiry date. Nor that human-written content is automatically more durable. Plenty of AI writing is excellent, and plenty of human writing is dreadful.

But I do think there's a new category of risk worth considering: stylistic obsolescence caused by the rapid evolution of AI-generated language.

The faster and more cheaply we produce content, the easier it becomes to underestimate the cost of maintaining it.

That cost isn't necessarily a loss of search rankings or measurable performance. It may be something subtler: a gradual erosion of the impression your brand makes on the people who encounter it.

And unlike an avocado bathroom, outdated writing doesn't announce its age with a particularly unfortunate shade of green.

The importance of having a voice

The obvious response is to tell AI to stop sounding like AI.

You can already ask models to avoid particular words, shorten sentences, remove em dashes or eliminate clichés. You can give them examples of writing you admire, provide detailed style guides and insist on a more natural tone.

These techniques help, sometimes enormously.

But I think they address the symptoms rather than the underlying issue.

A business shouldn't be trying merely to avoid the stylistic habits of the latest language model. It should be trying to develop and preserve a voice of its own.

What does your organisation actually sound like? Is it serious or playful? Direct or diplomatic? Authoritative or inquisitive? Does it use humour? Does it make bold claims or prefer understatement?

More importantly, do those choices reflect something genuine about the organisation?

An effective brand voice is more than a collection of preferred adjectives and punctuation rules. It's an expression of personality, values and perspective.

And those are things that AI can help articulate, but can't independently supply.

A language model can imitate confidence. It doesn't know whether your business has earned that confidence. It can write enthusiastically about your commitment to customer service. It doesn't know what happens when a customer phones your office with a problem.

Those distinctions matter because humans are remarkably perceptive when it comes to authenticity.

We may not be able to identify precisely why a piece of writing feels wrong, but we can often sense a mismatch between what someone says and who we believe them to be.

A company that sounds like an excitable motivational speaker on its website but behaves like a conservative accountancy practice in real life creates a small but significant disconnect.

AI can make that problem worse by smoothing every business into a version of the same agreeable, articulate, relentlessly positive personality.

Equally, used well, it can help maintain a distinctive voice more consistently across a large organisation. It can help different writers work to shared principles, expose inconsistencies and adapt content without losing the character of the brand.

The difference lies in whether the organisation is directing the technology or allowing the technology to establish the voice by default.

Before you choose your next bathroom suite

So what should businesses do?

First, I'd encourage you to become more conscious of AI writing. Read your own website, your competitors' websites and the marketing emails arriving in your inbox. Look for the recurring patterns. Notice which expressions seem to be everywhere and which ones have already begun to feel dated.

See whether you agree with me. Perhaps you'll find these stylistic similarities perfectly acceptable. Perhaps you'll discover that they bother you more than you expected. Either response is useful.

Second, talk about it with other people in your organisation. Your colleagues may have very different attitudes towards AI writing. Some will prioritise productivity. Others will care deeply about originality or the impression your brand creates. Those are legitimate considerations that deserve an agreed approach rather than a series of individual decisions made by whoever happens to be producing content.

And finally, think seriously about voice.

Understand what makes your organisation sound like itself, and how AI can both support and undermine that identity. Recognise that people are extremely good at sensing when communication lacks authenticity, even when they can't explain precisely what's wrong with it.

AI writing is an extraordinary tool. Its capabilities will continue to improve, and I expect it to play an increasingly important role in the way businesses communicate.

But speed, fluency and convenience aren't the only measures of good writing. Longevity matters too.

The best business writing should have a character that belongs to the business, rather than to whichever language model happened to be fashionable when the content was created.

After all, you want your words to remain recognisably yours long after everyone has forgotten what colour bathroom suites were popular in 2026.

Andy here. Thank you, ChatGPT. Read here for my original version of this piece.