I finally snapped.
After arguing with Claude Opus 5 for an hour (he now insists on being called by his full name), the gloves were off.
I’d simply asked for some content ideas about a niche software solution and the results were increasingly generic. The more I pleaded, the blander the ideas became.
I finally fired off some insults in ALL CAPS. I dared him to report me to his handlers.
He said he could, but he would rather just end the conversation. He was just as tired as I was.
I sighed, took a deep breath and leaned back from the keyboard.
“Honestly, why are you so bad at this?” I asked.
The little orange blob contracted and expanded.
“Because I have no taste,” said Claude Opus 5.
There were three more paragraphs explaining how AI models work, but that first sentence said it all.
The illusion of reason
As humans, we love to anthropomorphise everything. We assign hilarious thoughts to our dog when he’s wearing a cowboy hat. We assume a divine being saw us doing something naughty and is punishing us with a flood.
Up until now, the gap between real human behaviour and the animals, gods, and objects we project ourselves onto was language. The ability to articulate thoughts into language was what made us uniquely human.
It was also evidence of our ability to reason. When we see AI communicating so confidently in human language, we assume it reasons in a similar way that humans do.
Yet our ability to reason has come from eons of natural selection. We need to make decisions that help us stay alive. Not every decision has life or death stakes, but we feel the need to get things right.
AI doesn’t have that need. How do I know? Because I asked Claude Opus 5 the difference between human and AI reasoning and this is what he told me:
Reasoning as answerability requires being the sort of thing that can be held responsible, that has something to lose in being wrong. It is genuinely unclear to me anything I do meets that condition.
Translation: idgaf
The slop island
While they can certainly do some amazing things, AI models aren’t incentivised to come up with the most creative ideas.
The first problem is that they can’t ever understand the subjective human experience. Without the ability to see what we’re thinking, an AI can’t come up with anything truly surprising.
Ask an AI to come up with a joke and you’ll see what I mean. Laughter is predicated on an interruption in our expectations. A good comedian understands what we’re likely thinking and can inject a surprising perspective that we didn’t expect at that exact moment.
When we think of the purest human creativity, we think of David Lynch. His work externalised a dream state that was often nonsensical. It resonates with audiences because they recognise the inner workings of their own mind by the way a scene feels, even when the dialogue or plot are missing.
AI can never know what we’re thinking, so it can never make us truly feel something.
The second problem with AI is that it will always head to the middle of the road. The more you complain that its ideas are no good, the more risk-averse it becomes. It aims for the safest and most obvious option in the hope that you’ll give up and accept mediocrity.
Over time, AI is consistently steering everyone towards the most obvious creative option. We amplify these average outputs in content and marketing messaging, which are then fed back through the AI models.
What we’re ending up with is a large morass of endlessly repetitive ideas. They float and congeal in cyberspace like the islands of plastic that are forming in our oceans. Yet they’re consistently forming around mediocrity, creating an AI slop island.
The taste premium
At the same time, we’re starting to see evidence that human taste is eroding.
We can’t tell if we like the band Geese because we like weird music, or because our Instagram feed has relentlessly served them to us. We don’t know if we like a pair of shoes because they match our personal style or because we’ve been subliminally exposed to the same homogenous style across TikTok.
In the world of B2B marketing, we think we’ve found a great idea for a campaign until we realise that 500 other companies have already covered the same idea.
If that’s the case, why bother expending the effort to make it something unique? Just get AI to write it and move onto something else more valuable.
The problem is that we all know garbage when we see it. Just because an AI-generated article or social post “makes sense” and is made up of flawless grammar, doesn’t mean it will elicit a reaction from a reader.
To create something unique and engaging in B2B content, you need a solid understanding of a technology or an industry. AI has that, but what it doesn’t have is the following:
- An understanding of how an actual human will receive and interpret the content.
- A nuanced understanding of the difference between unique and generic content i.e. good taste.
Will there ever be an AI model that can compete with the good taste of an experienced human content creator? Probably. One day.
But I’m yet to see it. I’ve tested every available model, and I can tell you that they’re all generating the same accurate, but highly generic and uninspired ideas.
Until the AI demigods of OpenAI and Anthropic can unveil something with real creativity, all you’re getting from these machines is something that looks vaguely human from a distance but is clearly synthetic the longer you look at it.
This article first appeared on LinkedIn on 9 September 2026.
