There’s a particular kind of hunger that comes from writing, the kind that has nothing to do with food. I’ve heard of writers who lock themselves in a room and refuse to eat or drink until they reach whatever finishing point they’d set for themselves. I’ve never gone that far, and I can’t fully imagine what it feels like to want something that badly. But I know a smaller, tamer version of it, the moment where you can’t leave your seat because if you stand up now, the thought you’re chasing will get away from you. It’s worth asking directly whether AI makes you a worse writer, or just a faster one, because that hunger is exactly what’s at stake.
That feeling comes from the same place every piece of writing comes from: a blank page and no shortcuts. You start with a few hundred words. If you keep going, you get a few pages. Keep at it long enough and, somehow, you have a few hundred pages. Nobody skips a step in it, not even the people who write fast.
AI can now hand you a few hundred pages before you’ve written a single word of your own. And I think a lot of people are quietly forgetting what the hunger felt like in the first place.
Why not knowing the ending is the whole point
Akira Kurosawa talked about this in an interview, giving advice to young filmmakers. When you go mountain climbing, he said, the first thing you’re told is not to look at the peak. You keep your eyes on the ground in front of you and climb one step at a time. Look up too often, and you’ll get discouraged by how far you have left to go.
That’s what writing is, most of the time. You start without knowing exactly what the end looks like. You have an idea of it, maybe a strong one, but you can’t see it completely until you’ve climbed there. The not-knowing isn’t a flaw in the process. It’s the process.
What “knowing the ending” actually means
J.K. Rowling has talked about living with the Harry Potter characters for years while she wrote the series. She’s said that Hagrid was never really at risk of dying, even though he would have made narrative sense to kill off. The reason: she’d known since her first year of writing that the final book would end with Hagrid carrying Harry’s body out of the forest, a bookend to the opening scene where Hagrid carries baby Harry to the Dursleys. That image kept him safe.
This is worth sitting with, because it looks like a contradiction of the Kurosawa point and it isn’t. Rowling knew an image, not a manuscript. She still had to write every page between the beginning and that scene, one book at a time, discovering along the way who would live and who wouldn’t. Knowing where you’re headed and knowing how to get there are different kinds of knowledge. AI collapses the distance between them. That’s the part I want to get at.
Does AI actually make you a worse writer? What the research shows
This isn’t just a feeling. A few things are measurable.
Reading itself is dropping among young people. The National Literacy Trust’s 2025 survey, covered by the BBC, found that only about a third of 8-to 18-year-olds said they enjoyed reading in their free time, the lowest level the charity has recorded in twenty years and a steep decline from where it stood back in 2005. Numbers ticked up slightly in 2026, but they’re still well below where they were a decade ago.
Researchers studying AI and thinking draw a distinction that’s useful here: cognitive offloading versus cognitive outsourcing. Writing an outline before an essay is offloading. It frees up working memory so you can keep thinking. Asking AI to write the essay, or even just the outline, is something else. You’re not storing your thoughts externally anymore; you’re handing the thinking itself to something else. An MIT study that had people write essays using an LLM, a search engine, or no tool at all found weaker neural connectivity in the group that used the LLM. That’s not proof writing makes you smarter, but it’s a real signal that something changes in how the brain engages with the task.
Paul Graham wrote about this in an essay called “Writes and Write-Nots.” His argument is that writing forces you to think clearly, because you can’t write down what you don’t actually understand. As AI makes it easier to skip that step, he predicts a growing split between people who still write and people who don’t, and he says that split is really about thinking, not typing. As he put it, a world divided into writes and write-nots is more dangerous than it sounds.
The people choosing the hard way back
Here’s the part I find hopeful. There’s a small but real trend of people, especially younger readers, choosing physical books over ebooks, choosing print on purpose in a world where the digital version is cheaper and more convenient. Some of it is aesthetic, some of it is nostalgia. But a lot of the reasoning people give comes down to wanting something slower, something that resists being skimmed or sped through. It’s not really about AI. It’s a broader pushback against friction being removed from everything. But it rhymes with the argument here: some people are noticing what gets lost when the hard part disappears, and choosing the hard part back.
“People said this about calculators too”
Here’s the obvious objection, and it deserves a real answer instead of a dodge. Every tool that removed effort from a hard task got this same complaint. Calculators would ruin math. Spellcheck would ruin spelling. Google would ruin memory. GPS would ruin navigation. Mostly, none of it happened the way the doomsayers predicted, and it’s fair to ask why AI and writing should be any different.
Here’s why I think it is. A calculator doesn’t do your reasoning for you; it does the arithmetic after you’ve already decided what to calculate and why. Spellcheck fixes a typo in a sentence you wrote. GPS still requires you to know where you’re trying to go. Each of those tools sits downstream of a decision that stays yours. AI, used to generate a full draft from a prompt, sits upstream of that decision. It can do the reasoning, not just the execution. That’s a different kind of tool, and it’s fair to hold it to a different standard.
It’s also worth saying plainly: AI has given real writing ability to people who didn’t have it before, people who had ideas but not the craft, non-native speakers, people who’d been failed by how writing gets taught. For them, AI can work like the calculator did, opening a door rather than skipping a step they’d already earned the right to skip. This piece isn’t really about those people. It’s about what happens to someone who already had the skill, and quietly stops using it.
What to actually do with it
So here’s the line, as concretely as I can draw it. Use AI to research a claim, check a fact, push back on an argument you’ve already made, or fix a draft you’ve already written. Don’t use it to generate the first version of an idea that’s supposed to be yours, the outline, the argument, the opening paragraph you haven’t wrestled with yet. If you can’t explain your own point without pulling up the chat, you haven’t actually written it, you’ve approved something. That’s the test I try to use on myself, and I don’t always pass it.
Kurosawa’s advice wasn’t really about mountains. It was about tolerating not knowing, long enough to find out. That tolerance is a skill, and like most skills, it weakens when you stop using it. I don’t know what it feels like to lock myself in a room until I finish something. But I know what it feels like to lose a day’s work of thinking to a tool that got me to the end faster than I earned it, and I’d rather not make a habit of it.
