AI Is Quietly Colonizing How You Think

When I talk through rough ideas with a large language model, it repeats back a structured version of what I said. I read it and think, “Yes!. That’s exactly what I meant to say.” I feel understood. Validated. The AI perfectly captured my thinking.

But did it actually?

Did I actually think that before I read the response? Or did I recognize a plausible version of my unstructured thoughts and claim ownership after the fact?

I can’t always tell. Recognition sometimes feels like origination.

I often speak loosely into a voice note when I’m using AI. And it returns clean words and prose that sound intelligent. I see my ideas reflected back. The structure makes sense. The argument flows nicely. I experience the feeling of authorship. As John Nosta mentions in his recent post, "What once felt like the 'imperfectly imperfect' construct of thought can now feel more like a polished product with a shiny yet superficial reflection."

This feeling is recognition, not origination. I’m auditing the statistical predictions of this raw material coming from a technology I didn’t build. The sequencing, the framing, and the emphasis on certain points are all compositional micro-decisions by the AI model. The LLM made them. I approved them.

Approval feels like authorship, but it's’ not the same thing.

Colonization Happens Before You Write a Word

I use AI to think. I ask for strategic advice. I explore ideas through conversation and dialogue. I test the ideas by pushing back against model responses. I’ve promoted this as good practice when using AI. And this does still remain true. But if I am repeatedly exposing my reasoning through AI output, and absorbing what sounds correct and discarding what doesn’t resonate, then what is really happening?

Over time, it’s possible that my own sense of “what sounds right” is being trained by these interactions. Not consciously. Not through a single interaction. It occurs through thousands of small decisions and calibrations across months and even years of daily use.

This is internalized homogenization. It’s not the obvious offloading problem, where everyone’s output sounds moderately similar (though it still might). It’s my internal thinking that is converging with the AI distribution without my conscious awareness.

Colonization, while dark, is the right word for this. The LLM learns your language, codifies it, and then presents it back to you with modifications as the “polished” version. Over time, you start conforming to the colonial reasoning patterns and believing they are your own insights. “As the AI's outputs are reabsorbed into human discourse, they begin to shape users' own expression and reasoning, which in turn influences the data used to train future models (Sourati et al., 2026).”

Homogenization is part of a structurally reinforced cycle. The colonizer doesn’t announce itself. It integrates so seamlessly into your workflow that you mistake it for native thought. It doesn't just reshape your thinking once. It compounds.

A Thought Experiment

I keep coming back to a thought experiment, mostly because of what I’ve been observing on professional networks.

Put 50 consultants in a room who all use AI for strategy. Give them the same business problem. Each has to develop an independent analysis.

Twenty years ago, you’d probably get close to 50 different approaches and ideas. This is called variance. Different frameworks, different points of emphasis, and different blind spots. This variation would be a reflection of different reading histories, different styles of thinking, and professional or lived experiences.

Now? There’s remarkable convergence. No, they didn’t copy each other. They didn’t become less intelligent. Over time, they all internalized reasoning patterns from the same AI model distribution. It’s possible that the quality of the average analysis might be higher. But the variance doesn't exist. And the variance is where breakthroughs are. It’s where the economic premium lives. It's the approach that seems wrong at first but then proves right.

Artificial Intelligence Essential Reads

The Thought That Didn't Happen

Who Is Training Your Character?

When everyone’s reasoning converges towards the same insights, the outliers get trimmed before they get a chance to bloom. The LLM has effectively shaped their reasoning.

Expertise Fallacy

I expect a lot of pushback on these claims. There will be those that argue this is no different than any technology. The offloading frees up space for something better. The homogenization “doesn’t apply to me”. They’ll be those who argue that novices can’t verify the AI output correctly but experts can. Experts have domain knowledge and expertise to catch errors, evaluate themselves, and reject bad suggestions. I’ve made this argument myself. I’ve even published versions of it.

But the expert safeguard has a hole in it. When AI output aligns with their sense of how the problem should be analyzed, they don’t notice when their own sense of “how these problems should be analyzed” has already shifted towards the model’s reasoning. The verification process feels rigorous, but they are increasingly operating in a narrow window without realizing it.

A recent report by Sourati and colleagues (2026) found that, “rather than actively steering AI output, users often defer to mode suggested continuations. They select options that seem "good enough" instead of generating their own.” Building on this, I would argue that, "good enough" is indistinguishable in our brains from "correct" which is precisely why the shift goes undetected.

The Internal Homogenization of Thought

This is the colonial mechanism at the cognitive level. The imported structure becomes invisible over time and begins to feel native. You stop being able to tell the difference between "I think this is right" and "this matches the pattern I've been trained to recognize as right."

I started this post by questioning whether my ideas are actually mine. I still don’t have a confident answer.

I do have a suspicion. The more fluently someone uses AI to "think" and "reason", the harder it becomes to locate who is thinking. And the people most at risk are the ones who believe they’re too expert to be influenced. This includes me.

The questions I keep asking myself are, “Am I one of them? Have I been colonized so gradually that I can’t tell the difference?” I don’t know, but the inability to answer that question bothers me, and it may bother you as well.