Shaming the customer is not a business model

The anxiety behind the backlash is entirely reasonable. The marketing strategy it has produced is not — and it does most of its damage to people who were never part of the argument.

I keep seeing a particular kind of post.

A writer points out the obvious signs that somebody used ChatGPT. A designer lists the visual conventions that reveal a website was built with Claude. A consultant explains why AI cannot possibly reproduce the value of their experience.

The details change, but the argument is the same: I can tell you used AI, the result is inferior, and serious people will judge you for it.

These posts do well. They attract long comment threads from people in the same profession, agreeing that the work is slop and reassuring each other that genuine expertise will win in the end.

I should say immediately that I’m not a neutral observer of this, and not because I’m above it. I’m a writer. I’ve done this work for fifteen years, and the same technology is putting pressure on it. I built my own website with Claude, so I’m one of the people being described in those posts. And I’m developing a methodology whose central claim is that judgement matters more than production — which means “the backlash is anxiety rather than analysis” is an extremely convenient thing for me to conclude.

Hold that against what follows. I’ve tried to be fair to the anxiety, because it’s the correct response to the situation.

The threat is real, and the analysis is often right

People have spent years learning to write, design, advise or create. AI has arrived and made parts of that much easier to produce — in some cases producing, in minutes, something uncomfortably close to work people previously charged thousands of pounds for. That isn’t an abstract debate about quality. It’s an economic threat, and treating it as a failure of attitude would be its own kind of contempt.

The observations are frequently accurate too. ChatGPT has recognisable habits, Claude-built sites converge on similar layouts and typography, AI images have their own tells. Anybody who says they can often spot it is usually telling the truth.

The problem isn’t the diagnosis. It’s what gets built on top of it. If somebody accepts a first output without questioning it the result will feel generic, but that is a criticism of a weak process rather than proof the tool can’t produce anything worthwhile — and the weakest use is routinely presented as the only one. Which is convenient, because it supports a comfortable conclusion: the tool is bad, the customer was wrong to use it, and the old way of buying the service remains the only credible option.

“Slop” stops being a description at that point. It becomes a way of frightening people back into the market.

The part I find hardest to defend

One post I saw recently argued that an obviously Claude-built website showed its owner couldn’t be bothered to put effort in.

I’m not going to name them. This is their living, and the pressure behind the argument is real.

But look at what’s being said. Somebody has built a working website for their business. It may not be original, but it explains what they do and gives potential customers somewhere to find them. Before generative AI they might have had no website at all, or bought a cheap template that was every bit as generic.

Why does using an accessible tool demonstrate a lack of effort? And why would publicly embarrassing that person make them more likely to value professional design?

This is the contradiction at the centre of a lot of anti-AI marketing. It tries to prove the value of human expertise by showing contempt for the people who might need it.

The same sentence, aimed at two very different people

That website owner is worth staying with, because they expose something the argument never addresses.

The slop debate has a cast, and it is a small one. On one side, executives generating thought leadership they haven’t thought about. On the other, professional writers pointing out that they can tell. Both are people for whom publishing was already available — a communications team, a ghostwriter, a byline somewhere, a habit of having opinions in public. The largest group of new users doesn’t appear in the argument at all.

Somebody runs a plumbing firm and knows precisely why most bathroom leaks are a failure of sequencing rather than materials. Somebody keeps the books for forty small companies and can tell within a fortnight which ones are about to have a cash flow problem. Somebody has run a restaurant for eleven years and understands supplier pricing in a way no trade publication has ever bothered to write down.

None of them knows how to structure an article. That is not a gap in their expertise. It is a gap in an unrelated skill that has, until recently, been the price of entry.

“I can tell you used AI” does completely different work depending on which of these people it lands on. Aimed at an executive with a communications team, it’s a reasonable complaint about somebody outsourcing a thought they were being paid to have. Aimed at the bookkeeper, it’s a criticism of somebody for not having a skill nobody was ever going to give them.

The posts don’t distinguish. That’s the part I keep returning to, because a criticism genuinely about standards would need to — and this one never does.

The knowledge was there. The writing wasn’t.

Before generative AI, the bookkeeper had three options, and it is worth being honest about how poor all three were.

They could try to write it themselves, which for most people means several evenings, a result they don’t like, and a growing suspicion that they aren’t the sort of person who does this. They could pay a professional, which is straightforward advice from anybody who has never had to justify £600 for an article to a business turning over £200,000. Or they could publish nothing, which is what nearly all of them did.

The third option was so overwhelmingly the normal one that we stopped noticing it was a choice. An enormous amount of practical, specific, hard-won knowledge simply stayed inside businesses, because the people holding it couldn’t perform the genre that would have let it out.

I’ve spent a lot of this project arguing that organisations don’t have a content problem so much as an account-of-themselves problem — that the evidence is usually present and the interpretation is missing. The small-business version is simpler and more literal. The knowledge was there. The sentence-making wasn’t.

Where the gatekeeping claim holds, and where it doesn’t

So here is the sharper way to put it. When every use of these tools by an inexpert writer is dismissed as slop, the standard being defended is not quality. It is a hierarchy in which public communication belongs to trained writers, established executives, and organisations that can afford an agency.

I want to be careful, because “that’s just gatekeeping” is a lazy move that can wave away any standard at all, and I have spent a piece arguing that slop is real and predates AI entirely. A decade of search-optimised filler was written slowly and carefully by humans, and it was terrible.

The distinction I’d draw is between a criticism aimed at the work and a criticism aimed at the author’s standing. “This says nothing” is about the work. “You can tell they used AI” is about the author, and it only functions as a criticism if you accept that the tool disqualifies whatever came through it. Applied to a bookkeeper explaining a pattern they have watched for fifteen years, it defends nothing. The piece may be plain and structurally conventional. It may also contain the only useful thing published about that pattern all year.

Plenty of what small businesses publish will still be worthless, for a predictable reason: somebody has been told content marketing matters, has nothing particular to say, and now has a machine that will produce paragraphs about it. Volume was always the failure mode and the cost of volume has fallen to nothing. But what separates the two isn’t the tool or the polish. It’s whether the person brought something the model could not have supplied — a pattern they noticed, a number from their own books, a case where the standard advice failed. That is the same test I apply to my own use of AI: the tools are good at developing and stress-testing an idea, and useless at having noticed anything in the first place. A restaurant owner has fifteen years of noticing, and that was never the part that was missing.

Adequacy got cheaper. That’s the actual change.

The uncomfortable truth underneath all of this is that some production work is becoming cheap. Plenty of businesses don’t need an award-winning website but a credible place that explains what they do, don’t need every post handcrafted over several hours but a useful idea expressed clearly, don’t need a large agency process but one experienced person to help them make a few decisions and use the available tools well.

That doesn’t make expertise worthless. It means adequacy has become easier to achieve, and the opportunity for specialists sits above that line: original identity, user research and accessibility rather than a functioning page; reporting, evidence and argument rather than competent sentences; strategy connected to customers rather than more assets.

The alternative to contempt isn’t complicated. A designer could say: AI has made it easier to build something functional; I can take it beyond the defaults and make it distinctive, accessible and commercially effective. That concedes what’s changed and states plainly what expertise still adds. It’s just a harder sentence to write than “look at this rubbish”.

Where my argument is weakest

Three places, and I’d rather name them than have them found.

The economic version of the professionals’ case is stronger than the version they’re making. Cheap adequacy can destroy the market for good work regardless of whether it should. If most buyers can’t tell the difference and the adequate version costs nothing, the market for the better version can shrink to almost nothing too — and “move upmarket” is advice that works for some people and not for the profession as a whole. Everybody cannot move above the line. I don’t have an answer to that, and the posts I’m criticising are, in a clumsy way, reacting to something true.

Every technology has made the access promise. Blogging was going to democratise publishing. So was social media, and so was cheap video. Each time the barrier to producing fell, and each time the barrier to being read rose to compensate, because attention didn’t expand to match. The plumber may now write the article and be read by nobody, and I don’t have much of an answer beyond noting that not being read is a better position than not being able to write it down. “Latent expertise waiting to be released” is also a flattering story: some businesses published nothing because they had nothing to add, and the friction was doing a quiet filtering job we are now removing.

And the position costs me nothing. It’s easy to be relaxed about a plumber publishing. It would be harder to stay relaxed about the plumber’s article replacing a commission, and I haven’t been tested on that.

What the backlash is actually telling us

Its popularity is a signal in itself. It shows which professions feel most exposed, where people are struggling to explain their value in terms other than production, and how tempting it is to turn professional anxiety into moral judgement — that people using AI are lazy, their work is embarrassing, and discerning customers will eventually come back.

Some will.

But hoping customers abandon a useful tool isn’t a strategy, and treating every acceptable AI-assisted result as an insult to the craft mostly signals that the craft has stopped being able to describe itself.

If somebody can now explain their own business without paying us, the answer isn’t to tell them to stop publishing. It’s that what we’re selling has to be something other than access to words. That was always a strange thing to be charging for. It’s now an untenable one.