How AI turns your original ideas into conventional thinking
A friend sent me an article recently with a message that said, essentially: This sounds an awful lot like you.
She wasn’t wrong.
The article sounded familiar. It asserted that AI has made it possible for marketers to produce more content, faster, but that all this extra output creates work elsewhere: more reviewing, more approvals, more versions, more decisions, and more opportunities for a weak brief or broken process to slow everything down.
I have been making many of those arguments for the past year. In fact, less than two weeks earlier, I’d published an article arguing that AI might speed up production but doesn’t magically remove the real bottlenecks around workflow, approvals, coordination, and decision-making.
Naturally, my first reaction to this new article was: Hang on a minute. Then I started searching for related content. That’s when everything got a lot more interesting.
It’s not just one article
I found another recently published article by another author who made a remarkably similar argument: AI doesn’t fix broken marketing processes; it exposes them. Faster production can make approval bottlenecks worse, unclear ownership still creates delays, and producing more content doesn’t necessarily mean getting better marketing out the door.
Now I have three independently written articles, published within just a few weeks of each other, all circling the same territory.
That could, of course, mean we were all noticing the same thing at roughly the same time, and we’re all writing about that for respected industry publications. That happens.
But then I remembered something else. I had already finished my next article and sent it to the publisher. It hadn’t been published yet.
And that one was even closer to these other articles.
The article they couldn’t have seen
My next piece argues that AI creates the draft faster, but it doesn’t make the decisions better.
It covers weak briefs, approval bottlenecks, AI generating more options for people to disagree over, and extra content creating more material to review, revise, approve, reject, or rebuild. It looks at the danger of confusing productivity with progress, and why human judgment becomes more important as production becomes faster and cheaper.
The overlap is striking. But there’s a problem with the obvious explanation. I hadn’t read the first article that my friend brought to my attention, before I wrote mine. And, unless he has developed the ability to read unpublished Word documents on my computer, he hadn’t read mine either.
So maybe this isn’t an outbreak of plagiarism. Maybe nobody’s co-opting anybody else’s thoughts. Perhaps something else is going on, and maybe that’s both good and bad for establishing our AI knowledge base.
Read more: AI Made Marketing Faster Than it Made Marketing Better
Is AI the common denominator?
Like many marketers, I use AI to challenge my thinking, explore an argument, organise ideas, spot weaknesses, and help me sharpen what I’m trying to say.
I don’t begin with an empty prompt box and ask it to supply my opinions. My experiences, frameworks, observations, anecdotes, examples, and point of view come first. But AI is still part of the process. For me, it’s a second opinion that can reveal issues before I publish them.
I have no idea whether the other authors used AI while developing their pieces, and I’m not suggesting that they did. But it made me wonder what happens as more of us use LLMs to shape or expand our thinking.
If several people start with related questions and ask their LLMs to help explore them, how often will their models steer them toward similar arguments, examples, structures, or even rhetorical devices?
Ask an LLM why AI productivity hasn’t necessarily improved marketing, and it might produce the arguments I’ve mentioned here.
None of these ideas is wrong. Quite the opposite. They make perfect sense. And that might be exactly the problem.
I warned about this two years ago
The irony is that I’ve written about a similar concern already.
More than two years ago, in an article warning about ‘fauxthenticity’ invading email wisdom, I described three stages of thought leadership: regurgitation, adaptation, and innovation.
My concern then was that the internet was already awash with mediocre content masquerading as thought leadership. We could find plenty of regurgitation and adaptation, but considerably less genuine innovation.
That creates a problem with AI-powered searches.
LLMs learn from or retrieve existing content. But just because an idea appears thousands of times across the internet, that doesn’t make it good. It might simply mean thousands of people have repeated the same mediocre thinking.
This was true back in the days of organic and paid search, and even more so now with AI-generated results.
This worries me. If most of the content available to AI is derivative, superficial, or simply regurgitated conventional wisdom, how can the machine distinguish that from genuinely valuable thought leadership?
It can’t necessarily assume that the less common idea is the better one. Quite the opposite! The sheer volume of similar content can make the prevailing view look authoritative, even if it’s wrong.
Then AI produces more of it.
That AI-assisted content gets published and added to the information ecosystem, giving future AI systems even more material that reinforces the same thinking. Regurgitation begets regurgitation.
So, my concern wasn’t just that AI would create an echo chamber of bad information. It was that mediocre thinking could become self-reinforcing because there’s already so much of it for AI to learn from.
I still think that risk is very real.
But now I’m wondering whether I missed the other side of the problem: What happens when the original idea is good?
Read more: How AI Amplifies ‘fauxthenticity’
Good ideas can lose their provenance, too
We tend to talk about AI content problems in terms of hallucinations, misinformation, mediocre copy, and bland sameness. But an idea doesn’t have to be wrong to lose something in the AI ecosystem.
A genuinely useful idea gets published. It gets indexed, discussed, cited, summarised, quoted, adapted, and perhaps encountered by AI systems through training or retrieval. Eventually, it becomes easier for an AI assistant to deliver a version of that thinking when someone else asks a related question.
That person might never have heard of the original thinker. They might not even realise they are building on an established body of thought. The answer arrives looking logical, and it probably is.
But somewhere along the way, the connection between the idea and the person who developed it gets weaker.
The idea hasn’t necessarily been stolen. It has been normalised. And I know what that looks like, because I just saw it happen.
“Of course, we know holistic email marketing”
A few nights ago, I was having dinner with two email marketers from Lithuania. They had no idea who I was, other than my name was Kath.
At one point, I asked whether they had heard of holistic email marketing. Their response was immediate: “Of course!”
What they didn’t know was that I had originated the concept and introduced it to the industry in the 1990s. I’ve written about it for years, spoken about it on stages around the world, built methodologies around it, and eventually wrote a book called Holistic Email Marketing.
And do you know what? I loved their response.
Because, in one sense, that is what you want when you introduce an idea. You want it to spread, you want people to use it, and you want it to become part of the industry’s language and knowledge.
Holistic email marketing has now become so normalised that two email marketers from another country can know exactly what I mean without knowing where the term came from.
That isn’t an AI problem. Ideas have always outgrown their originators.
But here’s the difference: Before AI, that process took years. AI can now compress that journey considerably.
From original thinking to conventional wisdom — faster
Before AI, ideas largely moved through people. Someone heard you speak or read your article. They mentioned the concept to a colleague. An agency adopted it. A trainer included it in a workshop. Someone wrote about it, and others built their conference presentations on it.
Over time, the idea became familiar, and its origins became less important than its usefulness. Now we have another layer in that chain.
An LLM can encounter an idea, connect it to adjacent ideas, reframe it, and surface it in response to a completely different person’s question. And it can do that at scale.
This potentially shortens the distance between “original idea” and “Everybody knows that.” For anyone whose business partly depends on original thinking, that creates an interesting problem:
What happens when your ideas become generic faster than your reputation can travel with them?
AI doesn’t have to copy your words to commoditise your thinking
Traditionally, when we think about plagiarism, we look for matching language. Did someone copy our sentences, examples, or frameworks?
But that isn’t necessarily how this works. An LLM doesn’t have to reproduce your paragraph word for word to carry the underlying idea forward.
Take this argument: AI makes production faster, but the real constraint is judgment.
That could reappear as “The bottleneck has moved from creation to decision-making.” Or “More content doesn’t create more value if your organisation can’t absorb it.” Or “The time you save in production gets lost further downstream.”
Different words. Same intellectual neighbourhood.
Once enough versions exist, the original thought starts blending into the background. It becomes part of the conventional wisdom that passes for industry knowledge.
But where did the LLM get the idea?
This is where things become even murkier. Here, are four ways an AI system might encounter someone’s thinking.
- Publicly available articles form part of the material used to train models.
- AI tools capable of searching or retrieving current information encounter a published article while researching a topic.
- A writer feeds an article into an AI tool and asks it to analyse or build upon it.
- An LLM arrives at a similar conclusion simply because the underlying logic exists across thousands of other sources.
The problem is that the person on the other end of the prompt probably doesn’t know which of these things happened. They ask a question. The AI proposes an argument. It sounds intelligent, so they develop it.
There’s no little footnote saying: By the way, this line of thinking owes something to an article somebody else wrote three years ago.
That makes provenance increasingly difficult to see.
Thought leaders want their ideas to spread
As a thought leader, I have to be careful not to become territorial. The whole point of thought leadership is to influence how other people think. If you develop an idea, and it changes the way an industry approaches a problem, that’s a success.
Holistic email marketing is a perfect example. I’m delighted that people use the concept or talk about holistic email marketing without prompting. My goal was never to stand beside every use of the phrase, waving my book and shouting, “That was me!”
But thought leadership also has a commercial imperative.
A recognisable body of thinking establishes expertise. Expertise creates opportunities. People invite you to speak because you have something distinctive to say. They hire you because you have developed valuable frameworks and thinking. They read your work because they associate particular ideas with you.
If AI accelerates the point at which distinctive ideas become generic industry wisdom, the useful lifespan of original thinking becomes shorter.
The innovator does the hard work. The industry gets the benefit. And AI can get us faster to the point where everyone forgets where the idea came from.
Perhaps this is the real AI sameness problem
We’ve talked endlessly about AI making copy sound the same: the same phrases, the same sentence structures, the same “It’s not X, it’s Y” constructions, and the same suspicious fondness for em dashes.
But now I think we’ve been looking too far downstream.
AI can also make the thinking process feel the same: the framing, the metaphors, the examples, the logical progression of an argument, and even our conclusions.
Not because people can’t generate original thought, but because increasingly we’re asking the same systems to help us develop that thought.
If everybody starts at the prompt box, everybody is drawing from roughly the same pool.
That means differentiation must happen before the prompt.
What have you observed that somebody else hasn’t? What have you learned from 20 years of doing the work? What have customers taught you? What have you tested? Where do you disagree with conventional wisdom? What framework have you developed? What connection have you made that isn’t obvious?
That’s where the value is.
AI can help you interrogate those ideas, sharpen them, and challenge them. But it shouldn’t be the place where all of your ideas originate.
How to preserve original thinking when using AI
I’m certainly not going to stop using AI. That would miss the point.
But we need to become much more deliberate about where originality enters the process.
Start with your own observations, experience, research, data, arguments, and frameworks. Keep track of when you first develop important ideas. Name distinctive frameworks where it makes sense, and link new thinking back to the work that preceded it.
When AI produces something that sounds wonderfully clever, challenge it. Where did this come from? Is it actually new? Am I adding something? Would I have arrived at this conclusion without the machine?
Because there’s a big difference between an idea that sounds intelligent and an idea that is genuinely yours.
The next stage of the AI echo chamber
Two years ago, I worried about fauxthenticity: that AI would learn from the sheer volume of mediocre, derivative content already masquerading as thought leadership, then produce even more of it and reinforce the same thinking. I still think that risk is very real.
But now I’m wondering about the other side of the problem: what happens when the idea AI picks up is genuinely original and good?
AI doesn’t need to steal an idea to change its value. It can potentially help spread that idea so quickly, in so many forms, and through so many voices that eventually nobody remembers it was an idea at all.
It simply becomes something everybody knows.
As the two Lithuanian email marketers reminded me over dinner, that process existed long before ChatGPT.
AI might simply be putting it on fast-forward.
The threat to thought leadership is not that AI steals your ideas. Instead, it could be that AI makes them everybody’s.