Joan Didion opens "The White Album" with a sentence about the stories people tell themselves. Its force comes from the essay around it: the observations, choices and uncertainty that give the line meaning. A language model can produce a fluent sentence in a similar style. That observation alone cannot tell us how a published essay was made or what its author intended.
This is the distinction that matters, and it is the distinction that our current conversation about AI and writing almost entirely misses. The question is not whether AI can generate text that looks like writing. It can. The question is whether the thing it generates is, in any meaningful sense, writing at all.
Voice Is Not Style
When we talk about what makes human writing irreplaceable, we often reach for the word "voice." But voice is frequently misunderstood as style - a set of aesthetic preferences, sentence patterns, and vocabulary choices that could, in theory, be catalogued and replicated. If voice were merely style, AI could learn it.
But voice is not style. Voice is the accumulated weight of every experience a writer has ever had, pressed into language. It is Didion's particular brand of California anxiety. It is James Baldwin's ability to hold rage and love in the same sentence. It is Ursula K. Le Guin's quiet insistence that imagining other worlds is a political act. These voices cannot be separated from the lives that produced them.
The Beautiful Imperfections
In "Politics and the English Language," George Orwell argues for choosing words to fit a meaning rather than relying on stale phrasing. His rules end with permission to break a rule when following it would produce an ugly result. That is a useful reminder that a writing checklist serves the writer's purpose, not the other way around.
AI can produce unexpected metaphors, uneven rhythms and other signs readers associate with a personal voice. Those features are not reliable tests of authorship. The more useful question is whether the writer can explain the choices behind a passage and show how it changed through drafting and revision. The Authors Guild's AI advocacy FAQ discusses the professional stakes for writers while acknowledging that generative tools can also be useful.
What Lived Experience Gives Us
Consider the difference between these two sentences: "The hospital smelled like antiseptic" and "The hospital smelled like the underside of a Band-Aid." The first is accurate. The second is human. It carries the specific, embodied memory of a child pressing a Band-Aid to their nose - a detail so particular that it could only come from a person who has been a child, who has been hurt, who has been comforted inadequately by adhesive strips.
This is what lived experience gives us: not information, but texture. Not facts, but the particular weight of having encountered those facts in a body, in a place, at a time. Writing that draws on lived experience doesn't just convey meaning - it conveys the experience of meaning. And that experience is, for now and possibly forever, something only a human can offer.
Writing as Resistance
In 2026, writing by hand has become a radical act. Not because it's efficient - it isn't - but because it is unmistakably, irrefutably human. The notebooks, the crossed-out lines, the coffee stains, the margin notes: these are the artifacts of a mind at work, and no amount of algorithmic sophistication can replicate them.
But the real resistance isn't in the medium. It's in the commitment to saying something true - something that arises from observation, from thought, from the specific perspective of a specific person trying to understand a specific moment. This has always been what writing is for. The algorithms haven't changed that. They've just reminded us why it matters.