I use AI every day, and I still say it falls short for writers. Those two things aren’t in tension. The tool is real and helpful for the routine parts of the job.
What it can’t do is the part that makes writing land. Pretending otherwise is how a writer ships competent, forgettable work and calls it finished. Who is that supposed to fool, the reader or the writer? AI is neither useless nor magic.
It works in a different mode from a human writer, and the difference shows up exactly where writing is supposed to matter most. That’s the flaw in the hype. The hype tells writers the hard part is solved, and they believe it until a reader closes the book halfway through and never picks up another one with their name on it.
Think of AI writing like a player piano. It reproduces the notes flawlessly, in time, without a wrong key, and for background music that’s useful. Nobody mistakes the player piano for the pianist who bends the tempo, leans on a phrase, and plays the same notes in a way that makes a room go quiet. The notes are identical. The thing that moves you is missing, and that missing thing is the whole difference between competent and unforgettable.
Why is writing a human activity AI can’t replicate?
Because writing is translating experience, emotion, and perception into language that makes another person feel something.
That translation draws on decades of lived experience, sensory memory, and emotional intelligence, and AI has none of them. AI predicts the statistically likely next word. A writer reaches for the true one. The two processes look similar only from the outside.
The gap shows up most sharply where the stakes are highest.
AI can produce a technically competent sentence about grief. A writer who has buried a parent can produce a sentence about grief that stops a reader cold. The distance between a competent sentence and one a reader carries for years is the distance between pattern-matching and lived experience. Maya Angelou described the agony of carrying an untold story inside you. (More on Maya Angelou Day) That pressure, the need to say something that matters and can’t be faked, is where the best writing comes from, and no model has ever carried a story it couldn’t bear to keep.
AI has no untold stories. It’s training data, and training data doesn’t ache to be said.
Watch: The Limitations of AI Writing (And What Comes Next) #CreativeWriting #FutureOfWriting #Writer, The Hidden History
Why is cultural context still a blind spot for AI?
Because culture lives in the spaces between words: idioms, regional expressions, generational references, and shades of tone that shift meaning depending on who speaks and who listens. A human writer handles these without thinking after a lifetime swimming in them.
AI approximates from patterns, and the approximation breaks down constantly. It can spot that a phrase is an idiom and still miss the cultural weight behind it, flattening a sentence with a double meaning rooted in regional history into something literal and lifeless.
The distance between a competent sentence and one a reader carries for years is the distance between pattern-matching and lived experience.
Research on AI translation keeps finding culturally specific phrases getting mangled. Translation is a clean test case because it strips the problem to one question: does the system understand what words mean in context instead of in a dictionary? The answer is reliably no.
For writers working across cultures, or writing characters from backgrounds unlike their own, getting this wrong produces writing that offends, alienates, or erases people who already get flattened often enough. I think a writer who hands that job to a machine and publishes the result has walked away from the responsibility that came with choosing the subject.
Why isn’t AI’s logic the same as creativity?
Because AI runs on pattern recognition: given input A, produce the statistically likely output B. Creative writing runs on the opposite principle.
The best writing surprises. It violates expectations, defies patterns, and makes meaning through juxtaposition and contradiction. The magic in a story is the moment when something happens you didn’t see coming and instantly recognize as true. That leap is the one thing pattern-matching can’t make.
Look at how the writers who last work. Rowling broke every rule of realistic fiction and made the result feel more real than reality. King builds dread by dropping ordinary people into impossible situations and letting them react the way real people do: badly, selfishly, and with desperate courage.
AI can produce competent plot summaries and outlines that follow standard structure. It can’t make the odd creative leap that turns a competent story into a memorable one.
It can’t decide to kill a beloved character in chapter three because the story needs that wound. It can’t see that the most powerful thing in a scene is what goes unsaid. The result reads like a summary of a story instead of a story.
Why does ghostwriting expose AI’s core problem?
Because ghostwriting takes the skills AI lacks most. A ghostwriter thinks in someone else’s patterns instead of writing in their style. You absorb another person’s worldview, speech rhythms, emotional triggers, and blind spots, then translate all of it into prose that reads as if the client wrote it. That takes emotional intelligence at a level AI can’t touch.
When a client describes the worst day of their life, a human ghostwriter reads the pauses, the deflections, the places the voice drops or speeds up. Then they judge which details serve the story and which would feel exploitative. That judgment is the part of my job I’d never hand to software. Somebody trusting me with the worst day of their life deserves a person who can tell when they’ve said enough, and when a detail would hurt people the book was never meant to hurt.
AI can mimic a style given enough samples.
Feed it ten chapters and it produces something that resembles the voice on the surface. The resemblance stops there. The odd word choices, the shifts in emotional temperature, the way a person sounds different talking about something that matters than performing expertise: AI flattens all of it into one approximation.
For memoirs, personal narratives, and legacy books, whose whole point is capturing something authentically human, that’s a deal-breaker, since an AI approximation defeats the purpose, and that’s why the uncanny flatness of AI voice is so easy to feel and so hard to fix.
What is the AI over-reliance trap for writers?
The more writers lean on AI, the more their own skills atrophy.
It’s the pattern that appears whenever a tool automates a skill. The people who depend on it gradually lose the ability to do the work without it. Mark Twain drew the line between the almost-right word and the right word, calling it the difference between the lightning bug and the lightning. Finding the right word takes an ear trained by years of reading and writing.
AI doesn’t develop that ear. It calculates probability. Writers who outsource word choice stop building the instinct that separates adequate from excellent, and they rarely notice it going. That scares me more than anything the machines do on their own, because the ear takes years to train and goes dull much faster once nobody uses it.
In May 2025, the Chicago Sun-Times and The Philadelphia Inquirer ran a syndicated summer supplement called Heat Index with a reading list in it. The list credited Isabel Allende with a novel called Tidewater Dreams and Andy Weir with one called The Last Algorithm. Neither book exists. According to Fortune, more than half the titles on the list were made up.
The writer, a Chicago freelancer named Marco Buscaglia, admitted he’d used AI and hadn’t checked what it gave him. He told Axios, “This time, I did not and I can’t believe I missed it because it’s so obvious.” King Features, which produced the supplement, said it has a strict policy against using AI to create material, and according to Fortune it terminated him.
I think that’s the over-reliance trap in one newspaper section. Buscaglia said he normally checked his material, and the time he didn’t, the machine invented books by living authors and the list went to print. I feel some sympathy for him, because the habit that burned him is the one I see writers building every day: trusting text because it sounds right.
Two more problems compound it. The first is homogenization.
When thousands of writers use the same tools, the output converges. AI text has fingerprints: the hedging, the parallel structure, the compulsive summarizing, the reflex to give three examples where one would do. The more everyone relies on it, the more all writing sounds the same and the less any voice stands out.
The second is the flood. AI generates content fast and cheap. That sounds like abundance and works like pollution. The signal-to-noise ratio drops, and good writing has to fight through an ocean of competent, forgettable material to find its audience. That’s the dynamic behind the curse of shallowness.
My take
Check everything the machine gives you. A newspaper reading list went to print with books that don’t exist because a writer trusted text that sounded right. The habit is spreading. Give AI the grunt work and keep choosing your own words, since the ear that finds the right one takes years to train and goes dull fast. Your ear is the part of the job no software replaces, so keep it sharp.
How is AI affecting entry-level writing careers?
Entry-level writers are getting squeezed. AI is eating the jobs where new writers used to learn: SEO articles, blog posts, product descriptions, social copy. That work paid badly and taught you to hit deadlines, write to a brief, match a brand voice, and handle revisions. Businesses that once hired a junior writer for blog content now run AI and have a senior editor clean it up. The junior position vanishes.
McKinsey research on automation has consistently found content-generation tasks among the most susceptible to displacement, and the trend accelerated once generative AI went mainstream.
Remove the bottom rung and the whole ladder to ghostwriting, longform, and book authoring is harder to climb. I think the companies cutting those junior jobs are saving a little money now and starving the pipeline they’ll need later. In ten years they’ll go looking for senior writers who can do what the machine can’t, and there’ll be fewer of them.
Nobody beats AI on price and speed, so out-producing it is a losing fight.
Out-thinking it is a fight writers can win. AI is a useful tool for brainstorming, editing passes, research, and rough drafts you then rewrite in your own voice, and those are legitimate uses that save real time. It still doesn’t think, feel, make creative leaps, capture an authentic voice, or make the editorial judgment calls that separate competent writing from work that matters.
The writers who do well will use AI as a tool while doubling down on exactly what it can’t replicate: emotional intelligence, cultural awareness, authentic voice, creative risk, and the willingness to put something personal on the page, the substance of using these tools well without letting them write for you. For the wider picture, the AI writing hub covers it, and if you want writing with a pulse no player piano can fake, that’s the work I do.
I’ve run out of patience for anyone selling AI as a replacement for writers. Use it for the chores and it’ll save you real time. Hand it the part the reader came for and you’ve given people who bought tickets for a pianist a player piano, and they’ll notice before the first chapter ends.
| Area | What AI does | What a human writer does |
|---|---|---|
| Emotion | Predicts the statistically likely next word | Reaches for the true one, drawing on lived experience |
| Cultural context | Approximates from patterns and misses the cultural weight of idioms | Handles idiom and tone after a lifetime immersed in them |
| Creativity | Produces competent summaries that follow standard structure | Makes the odd leap that surprises and rings true |
| Voice | Mimics surface style and flattens it into one approximation | Keeps the odd word choices and shifts in emotional temperature |
| Word choice | Calculates probability | Finds the right word with an ear trained over years |
| Facts | Can invent books that don’t exist | Checks everything the machine gives |
| Routine work | Brainstorming, editing passes, research, rough drafts | Rewrites those drafts in their own voice |
Over-reliance on AI is a slow loss you don’t feel going. I wrote The Death of Thinking about what outsourcing judgment costs writers, programmers and analysts.
Frequently Asked Questions
- Major newspapers ran a summer reading list. AI made up book titles., The Washington Post (via Boston.com)
- Chicago paper’s “reading list” full of fake, likely AI-generated titles, Axios
- Quote Origin: The Difference Between the Almost Right Word and the Right Word Is Really a Large Matter, Quote Investigator
