The Stigma Migrates. The Sameness Stays.

The warnings about generic AI content have become their own content category. Depending on your feed you might get the version about credibility, or the one about customers spotting AI images instantly, or the carousel explaining that the paint splashes and moody sunsets are making every business look identical. The advice at the end is always to be original.

Fair enough, as far as it goes. The AI look is real and it is everywhere.

But scroll past the warning and look at what surrounds it. The website of the business posting it was probably built on one of four or five platforms, from a template shared with a few thousand other businesses, with a hero image up top, service tiles in the middle, a testimonial slider and a newsletter box at the bottom. Their social graphics came out of a template library. Their welcome email sequence follows the structure everyone’s welcome email sequence follows, because everyone learned it from the same places. All of this sameness is at least a decade old and none of it has ever attracted a warning. There is no viral carousel about the dangers of the testimonial slider.

Sameness, on its own, has clearly never been the offence. Templates prove that daily. Whatever people are objecting to when they object to AI content, it is something else, and the research on what that something else is turns out to be genuinely awkward.

What happens when you label the same content two different ways.

Bynder ran a survey with two thousand consumers across the UK and US. Participants read two unlabelled articles on the same topic, one written by ChatGPT and one by a professional copywriter. Among people who preferred one over the other, 56 per cent picked the AI article as more engaging. Later in the study, once people knew which content was machine-written, 52 per cent said that knowledge made them less engaged with it. The words on the page never changed. What changed was the story people were told about where the words came from.

The Nuremberg Institute for Market Decisions found the same thing with advertising. Two groups saw identical ads. The group told the ads were AI-generated rated them as less natural and less useful. A field experiment by researchers at NYU and Emory measured the cost in behaviour rather than opinion: disclosing AI involvement in ad generation reduced advertising effectiveness by up to 31.5 per cent, for the same ads.

So when someone says an AI image looks soulless, there is a decent chance the soullessness arrived after the diagnosis, travelling backwards from the label to the picture. People believe they are judging the work. Mostly they are judging its biography.

There is older psychology here, and it predates AI by twenty years. In 2004, Justin Kruger and colleagues showed people a poem, a painting and a suit of armour, varying only how long the work supposedly took to make. Longer meant better. People rated identical objects as higher quality and valued them higher when they believed more labour had gone in, and the effect grew strongest when they had no other way to judge.

Most people looking at a small business online have no other way to judge. They are not designers. So they judge the labour, and this is where the template gets its exemption, because a template still smells of labour. Somebody chose it. Somebody dragged their own photos into it and stayed up rearranging the sections. Small effort, but human effort, and the viewer can feel it in there somewhere. AI output presents as effort removed, and later research found we score these differently by source: slow, effortful work from a human reads as care, while the same signals from an algorithm read as nothing, or count against it.

A 2023 replication of the Kruger studies found the effect held for quality and liking but came apart on monetary value, so treat it as a tendency with limits. A tendency is all the argument needs. It explains why customising a template redeems it in a way that no amount of careful prompting has ever redeemed an AI image for someone who has already decided what AI images mean.

The forty-year history of the current panic.

In 1985, Aldus PageMaker put professional page layout onto a personal computer. Anyone could suddenly produce publication-quality material without typesetting equipment or trained production staff. The professions reacted the way professions do. Designers feared their careers were about to become irrelevant now that the tools were available to everyone. The professional verdict on the output was that desktop publishing had flooded the world with badly designed documents, assembled by amateurs who put the wrong typefaces together and mistook owning the software for having the skill.

Read those complaints next to this month’s LinkedIn feed and the nouns are the only difference.

Desktop publishing eventually stopped being an argument and became plumbing. The complaint needed somewhere to live, so it moved into clip art. When clip art died it moved into stock photography, the era of the gleaming handshake and the woman laughing alone with salad, mocked by the same industry that bought the licences. After that, website themes. Then the drag-and-drop builders, which were the end of proper web design for about five years, until they became how web design mostly gets delivered, at which point the complaint packed again and moved into AI, where it lives now, frequently expressed through content produced with the previous tenant.

Forty years of this, and the sameness itself never once went away. Clip-art-era marketing looked like other clip-art-era marketing. Stock-photo websites looked like other stock-photo websites. Template sites look like template sites.

Meanwhile, in every one of those eras, some businesses were unmistakable anyway. You have met them. There is a business you follow whose site runs on an ordinary template, and you could identify their writing with the logo stripped off, because a specific person with a specific way of seeing their work is audible in every line. And somewhere in the same industry there is a competitor with a bespoke site, professionally designed, five figures spent, that you could not pick out of a lineup, because underneath the design budget nobody ever worked out what the business was. Forty years of tools have had opinions about both of these businesses and made no difference to either of them.

The difference between them is upstream of every tool either of them touched. One of them knows what it is. That knowledge survives cheap production, and its absence survives expensive production.

Which is why the two businesses above keep turning up in every era. One did the work underneath. The other keeps changing tools. If you have ever noticed that you are compelling for two hours over coffee and interchangeable on your own homepage, you already know which side of that line you are standing on, and you already suspect the next tool will not move you across it.

Questions this usually raises.

Do AI-generated images actually damage a brand?

The measurable damage attaches mainly to disclosure, and to the absence of anything underneath. Unlabelled, AI content often outperforms human content in testing. Labelled, identical content scores worse. A business with a defined voice using AI somewhere in its pipeline still reads as itself. A business with nothing to say reads as nothing, in any tool, and AI just produces the nothing faster.

Should I stop using templates?

No. Scaffolding is neutral. The question worth sitting with is whether anything recognisably yours stands on it. If your website could wear a competitor’s logo without anyone noticing, replacing the template will relocate that problem at some expense.

Can people really spot AI content?

Far less reliably than they claim. In Bynder’s research, only around half of consumers correctly identified which content was AI-written, close to a coin flip. What people detect at much better than chance is the absence of a point of view. The attribution to AI comes afterwards, and is often wrong about the tool while being right about the emptiness.

How does a small business stand out when everyone has the same tools?

Every competitor you have can buy your template, your platform, your image generator and your typeface this afternoon. None of them can buy your particular history, or the way you see the work, and most businesses have never put those into words even for themselves. The gap between what you know in conversation and what your business manages to say in public is where standing out actually lives, and closing it is identity work before it is marketing work. It is also the work most business owners have been avoiding for years, usually while changing tools.

The tools will keep turning over. The warnings will keep moving to whichever tool is newest, and each wave of warnings will be produced with the tools from the wave before. Somewhere under all of it sits the older question of what your business would have to say if saying something was the only way left to be recognised. Most businesses cannot answer it yet. The tool debate exists so they do not have to.