I'm quite glad the SEO slop era is ending

Slop didn’t arrive with ChatGPT. It just got faster, and started being made by something we’re allowed to be angry at.

I’m quite glad the SEO slop era is ending.

That may sound strange from someone who has spent most of his career writing for the web.

But I remember the period when a piece could be considered good content because it ranked.

Never mind whether the angle was original. Never mind whether it told the reader anything they hadn’t already seen elsewhere. Never mind whether an editor would have commissioned it without the keyword volume attached.

If it contained the right phrases, answered the right query and brought in traffic, it had value.

A lot of it was terrible.

Human-written terrible. Slowly researched, manually drafted, carefully keyworded terrible.

And because it performed, organisations produced more of it.

What that era actually cost

It helped turn editorial judgement into traffic acquisition.

Originality became harder to justify than search demand — you could defend a keyword with a volume figure, and you could not defend an interesting angle with anything. Journalism found itself competing with an industrial system designed to produce whatever Google was already telling everyone to look for.

I’ve argued separately that AI search risks repeating exactly this mistake with a new vocabulary. This piece is about something else: how we talk about the mistake we already made.

My stake in this, which isn’t neutral

I should be clear about where I’m standing, because Editorial Intelligence exists partly as a reaction to that era.

I spent years producing content at scale inside systems that rewarded volume and search demand over whether anything was worth saying. Not as a critic of the machine — as somebody operating it, reasonably well, for a long time.

The methodology is largely an attempt to build the opposite. Evidence before assertion. Arguments that have to be defensible. And, critically, the option of deciding something should not be published at all — which is the one decision the old system had no way of valuing.

That’s also why I’m wary rather than triumphant about what comes next.

The incentive that produced SEO slop has not gone anywhere. Volume was rewarded when it was expensive. It will be rewarded harder now it is close to free, unless somebody decides what else counts.

The historical amnesia

When I see another post lamenting the arrival of AI slop, I sometimes struggle with the memory loss.

Yes, there is AI slop everywhere. Yes, AI makes producing mediocre copy vastly easier.

I want to concede the strongest version of that objection rather than wave at it, because it is the serious one: scale is a difference in kind, not only degree. A hundred people producing bad articles is a nuisance. A system that can produce a hundred thousand is a different phenomenon, and the internet’s discovery mechanisms were not built for it. Anyone arguing that AI slop is merely the old problem with better tooling is not taking that seriously enough.

But slop did not arrive with ChatGPT.

We had SEO slop. We had content-farm slop. We had “10 reasons why” slop. We had 1,500-word articles whose main reason for existing was that somebody had identified a keyword with a respectable monthly search volume.

Every one of those was made by a person, deliberately, over hours.

Which raises the uncomfortable question underneath the current anger.

The fact that a human spent four hours writing something never automatically gave it editorial value.

If it had, the last decade of the web would have been much better than it was.

I suspect part of what makes AI slop feel worse is that it gives us something to be angry at. Human slop was produced by people doing their jobs inside incentive systems that rewarded it, which is a harder thing to be indignant about — because most of us were somewhere in that system.

What I’m actually against

None of this is a defence of slop. I’m completely against it.

I just don’t think authorship is what makes something slop.

The opposite of AI slop is not human writing. It is original insight.

Content becomes slop when it recycles ideas without adding evidence, synthesis or understanding. That can happen at four hours a page or four seconds a page. The slow version was more expensive and no better for it.

So the test isn’t who made it, or how long it took. It’s whether the thing should exist — whether it gives a reader something they could not already get somewhere else.

Effort was never a quality signal. It was just the only one we could see.

Where originality actually moves

This is also why I think the argument about AI and writing gets confused.

For B2B and informational publishing, I don’t think the highest form of professional skill is necessarily producing every sentence manually.

The scarce things increasingly look like judgement:

Finding an angle that hasn’t already been exhausted.

Knowing what evidence matters.

Understanding a customer well enough to say something useful.

Making connections between things everyone else has treated separately.

Knowing when the obvious brief is the wrong brief.

And knowing when the machine has produced something that sounds perfectly competent and says absolutely nothing.

That last one is the skill nobody lists, and it is the one that determines whether AI makes your output better or merely faster.

Originality moves into better questions, better arguments, better evidence and better editorial decisions. Not into prettier sentences.

An important boundary

I’m talking about B2B and informational publishing.

I don’t think you can automatically extend this argument to novels, music, film, illustration or any form where the human act of creation is part of what people are there for. Those are genuinely different questions, and the answers may well go the other way.

Writing a useful article about accounting software is not culturally equivalent to writing a novel. Somebody making a living from either deserves better than having the two collapsed together.

And I think treating every debate about AI-generated language as though those things are interchangeable makes the conversation considerably worse. It lets people defend the weakest possible case by pointing at the strongest, and it lets the other side do the reverse.

What I actually think

So no, I don’t particularly mourn the old content machine.

AI may be destroying one version of web writing. Parts of that version deserved to be destroyed, and were doing real damage to the parts worth keeping.

That isn’t a defence of what replaces it. A worse thing can follow a bad thing, and pointing at the bad thing is not an argument that the worse thing is fine.

But it does change the question.

The interesting one isn’t whether AI-produced copy is worse than human-produced copy. Plenty of both is bad, and the authorship never settled it.

It’s what we decide has value once producing competent copy is no longer scarce.

We had an answer to that for about fifteen years. It was traffic. It was not a good answer, and we’re only losing it now because something arrived that could exploit it faster than we could.