Personally, I don’t care whether you wrote what I’m reading. I couldn’t care less if you typed every character, dictated it, worked with a ghostwriter or used an LLM. What I do care about is whether it represents what you think and whether you’re willing to stand behind it.
There is an astonishing amount of AI slop out there. Effortless text, generated in seconds and published without scrutiny, repeats the same phrases and mannerisms. Once you start noticing it, you see it everywhere. AI made producing text cheap, and before you know it, you’re adding more slop to an already overly slopped world. But we’re drawing the wrong conclusion from that.
AI can produce bad writing extraordinarily efficiently. That doesn’t make using AI to write inherently bad. We’ve been through a surprisingly similar discussion with code.
Did you code it?
As AI got better at generating code, we started asking whether coding still mattered. That question focuses on the activity while most of us care about the outcome. We want to build something that solves a problem, and code happens to be the material from which we build it.
Sometimes working directly with that material is the point. There is craft in writing elegant code, just as there is craft in woodworking. But not every piece of code needs to express craftsmanship. Sometimes you just need the application to work.
Suppose you describe what you want to an AI coding agent, challenge its implementation and test the result. Along the way, you discover something you hadn’t considered, correct the agent and change your mind. Eventually, you ship a working application. Did you code it? Maybe not. Did you build it? Yes. Writing deserves the same distinction.
Did you write it?
Sometimes the words are the product. In literature or a beautifully written essay, the precise choice and arrangement of words are an essential part of what you’re consuming. That’s craft, and I have no interest in arguing against it. I love language.
A lot of writing has another purpose. You have a thought and want to convey it to someone else, so prose becomes the medium. You still want it to be clear and pleasant to read, but the reason for writing is the thought it carries.
Suppose you explain that thought to an LLM and argue with it. You reject some of its arguments, adopt others and discover holes in your reasoning. Eventually, you arrive at prose that accurately represents what you think. Does it matter whether you personally typed every word? I’m increasingly convinced that it doesn’t. The more useful question is: is this what you think?
Writing is thinking
One objection to AI-assisted writing is that writing is thinking rather than transcription. I agree. Countless times, I’ve discovered that I don’t quite know what I think until I try to put it into words. Explaining something exposes contradictions, and a sentence that refuses to work may reveal that the thought behind it doesn’t work either.
Typing is one way to externalize thought. Conversation does it too. You can start with an observation, explain it to someone and have them challenge it. In defending your position, you may realize they’re right, reverse course or try an analogy that turns out to break. You end up somewhere you couldn’t have reached when you started. That’s thinking. It doesn’t stop being thinking because the other side of the conversation is an LLM.
Mine doesn’t mean original
The idea doesn’t need to originate with me either. Very few ideas do. We read and talk. We borrow ideas from different places and combine them. Someone says something that changes how we think, and months later, that idea has become part of how we see the world.
AI doesn’t fundamentally change that process. It might introduce an argument I hadn’t considered. I can reject it or think about it and decide that it’s right. At that point, asking who originated the thought tells you less than asking whether I’ve made it my own and can defend it.
Authorship, to me, is therefore increasingly about accountability rather than provenance. If an AI coding agent generates a function and I ship it to production, I don’t get to shrug when it breaks and say that the AI wrote it. I shipped it. I own it. Why should writing be different?
Difficulty proves very little
With AI, we sometimes measure authorship by the visible effort behind the final artifact. That doesn’t make sense. Suppose I spend months developing a system that understands how I communicate and challenges weak arguments. It also recognizes when something doesn’t sound like me. After enough iteration, it can turn my thoughts into prose that I’m happy to put my name under with little additional editing. Why should I deliberately make the next article harder to produce? The whole point of building a good system is to benefit from it repeatedly. Difficulty is no evidence of authorship, and effort tells you little about value.
I can spend three days writing something that isn’t worth your time. Someone else can produce something in thirty minutes that changes how you think. You don’t owe me extra attention because I suffered more.
More people can contribute their thinking
Not everyone is a good writer. Some people struggle to structure their thoughts on a page, while others communicate better through conversation. Some write in their second or third language. And none of that makes their ideas less worth hearing.
For a long time, contributing an idea through writing has been partly gated by the ability to produce good prose. People with access to editors or simply a helpful colleague had an unfair advantage. AI levels the field, and makes similar assistance available to far more people.
If someone has spent twenty years developing expertise but struggles to write a compelling article about it, and an LLM helps turn that knowledge into something you can understand, you’re richer for having read it. They don’t need to become a wordsmith before they can contribute the thought.
Don’t make more slop
None of this excuses slop. AI gives us an astonishing ability to produce text, which creates an obligation to exercise judgment over what we publish. Just because you can generate another article doesn’t mean that you should.
Generating something costs you almost nothing. Reading it still costs someone else time. If you save yourself an hour by dumping unedited AI output onto the internet, you haven’t eliminated the work. You’ve merely transferred it to everyone unfortunate enough to read it.
Use AI. Build systems and make writing easier. Let it help you find words you couldn’t find yourself, but do the work that matters. Ask whether you have something to say. Challenge the argument and remove the crap. Check the facts, then decide whether the result deserves someone else’s attention. If you’re going to put your name underneath it, own it.
Why read anyone anymore?
You don’t have to. We have access to an extraordinary amount of knowledge, and if you have a question, you can ask an LLM. I do it all the time. But one answer is never every answer.
Ask a model a question and you’ll get a response shaped by what the model considers appropriate. It might be excellent and expose you to ideas you hadn’t considered, but it still represents one response. People bring different experiences and disagree. Sometimes they’re wrong in interesting ways. That’s worth preserving. An idea doesn’t need to have never existed before for someone’s perspective on it to be worth hearing. They need to have thought about it and be willing to stand behind it.
The question who wrote this? is becoming less useful than it once was. It bundles the origin of the idea with production and editorial control. It also hides the most important part: who ultimately stands behind the words. AI pulls those roles apart.
When you publish something, ask whether you’ve thought about it and whether the words represent what you mean. Are you willing to be accountable for putting them in front of others? If so, they’re yours. We should care less whether someone wrote every word. We should care very much whether the words are theirs.