Two things are true at once. AI tools have removed real barriers for writers with dyslexia, chronic illness, and no MFA Rolodex. The same tools are being used to submit machine-generated manuscripts to prizes under a single human name. Some writers deploy the accessibility argument honestly. Others deploy it cynically. The literary establishment has refused to do the work that would let anyone tell the difference. At YouWrite we sell a story generator, so we have skin in this fight. We would rather the line get drawn cleanly than keep getting blurred.
The access wins are not hypothetical
A writer with dyslexia who has spent twenty years being told their query letters are unreadable can now produce clean prose on the first pass. A caregiver with forty-five minute windows between tasks can dictate scene beats into a chatbot and come back to a draft instead of a blinking cursor. An ESL novelist who thinks in Tagalog or Yoruba can write the book they always had and then sand the English without paying a freelance editor four thousand dollars they do not have.
These are not edge cases. The Authors Guild's 2023 survey of working writers put median full-time author income at roughly $20,000. The MFA pipeline that still gates most debut fiction deals costs more than that per year at most programs. When critics treat any AI use as cheating, they defend a system that has always cheated people without money, time, or neurotypical reading speeds.
The access argument is the strongest thing AI advocates have. Which is exactly why it gets weaponized.
The conflation that broke NaNoWriMo
In September 2024, NaNoWriMo posted a statement saying that categorical opposition to AI carried, in their phrasing, classist and ableist undertones. The backlash was immediate. Multiple board members resigned. Sponsors pulled out. The organization walked the statement back and then announced in early 2025 that it was shutting down entirely, citing financial collapse.
Read the original statement carefully. It refused to distinguish between a dyslexic writer using Grammarly's generative features to clean a paragraph and a user pasting "write me a 50,000 word novel about a vampire detective" into ChatGPT and submitting the output. Both fell under "AI use," both got the same defense, and the defense leaned on disabled writers as human shields for the second behavior. That is what made people furious. Not the access argument. The conflation.
The same conflation now shows up at the prize level. In 2024, Granta and several Commonwealth-adjacent literary outlets were caught up in disputes over submissions reviewers suspected were largely AI-generated. Pangram Labs, an AI-detection startup, was reported to have retroactively screened past prize submissions and flagged a meaningful percentage as likely machine-written. Detection tools are unreliable enough that retroactive screening is its own ethics problem. The underlying anxiety is not invented. Editors are reading submissions that feel hollow in a specific new way, and they have no agreed-upon vocabulary for talking about it.
The line nobody wants to draw
Here is the distinction almost every public argument refuses to make.
Scaffolding is when AI helps a human author execute a vision the human is holding. The plot is yours. The characters are yours. The scene you are trying to write exists in your head before you open the tool. AI helps you get unstuck, fixes your grammar, suggests three ways a character might respond so you can reject all three and find the fourth, or translates your dialect-heavy draft into standard English without losing the voice. The judgment, taste, and intention stay with the human.
Ghostwriting-by-algorithm is when the AI is holding the vision. You prompted a premise. The model made the choices. You edited the output for typos and sent it out under your name as solo human work. The judgment lives in the model's training data and its statistical defaults. You are the typist and the submitter.
Most real use sits somewhere on that spectrum, not at one pole. A useful question to ask yourself: if the model vanished tomorrow and you had to defend every paragraph in a workshop, could you? Not whether you typed the words. Whether you chose them, in the sense of holding alternatives in your head and rejecting them for reasons.
If yes, you are scaffolding. Disclose the tool if the venue requires it and move on. If no, you are submitting work that is not solo human work, and saying so is not gatekeeping, it is accuracy.
What the institutions owe everyone
Prize committees, magazines, and writing organizations keep issuing AI policies that say things like "no AI-generated content" without defining generated. This is the root cause of both bad outcomes:
- Writers using AI as a wheelchair-equivalent accommodation get accused of fraud.
- Writers using AI as a ghostwriter slip through because nobody can prove what counts.
A workable policy needs a clear definition of what kind of AI assistance is permitted, ideally framed around who is making the creative choices. It needs disclosure categories granular enough that a dyslexic writer using sentence-level cleanup does not have to file the same form as someone who prompted a plot. And it needs to drop the fantasy that detection software can sort this out post hoc. Pangram, GPTZero, and Originality.ai all publish accuracy claims that collapse on writing by non-native English speakers and neurodivergent authors. Using them as evidence reproduces exactly the bias the access argument was trying to name.
None of this is hard. It is just unflattering. It requires admitting that "is this AI" is the wrong question and "who made the choices in this manuscript" is the right one.
Where YouWrite sits, honestly
Our story tool produces drafts. If you paste one into a prize submission and put only your name on it, you are doing the thing this post is about. We would prefer you use it to break through a stuck chapter, generate options you then cut apart, or get a complete bad draft you can rewrite into a good one. That is what it is good for. It is not good at taste, and it never will be, because taste is the part that has to be yours.
The writers most hurt by the current confusion are the ones the access argument was built to protect. Every prize jury that quietly assumes ESL syntax means GPT, every workshop that side-eyes a disabled writer's suspiciously clean prose, every magazine that bans "AI" without defining it, makes the field smaller and whiter and richer than it already was. Defending that outcome in the name of literary purity is its own kind of fraud.