Does Writing Make You a Storyteller?

AI is exposing an uncomfortable distinction between conceiving a story, writing its prose, and being recognized as its author.

Consider two aspiring novelists. The first has spent years developing a fictional world, working through its history, characters, conflicts, and the consequences of the decisions those characters make. They understand how the story should unfold, know what motivates their characters, and can explain why a particular scene belongs in the narrative. Unfortunately, their writing is dreadful. Their grammar is inconsistent, their sentences awkward, and their attempts at description frequently fail to communicate what they have imagined.

The second novelist has no such difficulty. Their prose is polished, their grammar impeccable, and their sentences flow naturally. They can describe a room beautifully, capture the sound of rain against a window, and turn an ordinary observation into something eloquent. Unfortunately, their characters lack motivation, the plot barely holds together, and the story has little reason to exist beyond the language used to tell it. Which one is the better writer? I suspect most people would hesitate before answering, because we recognize that imagination and technical proficiency are different abilities, although we've become accustomed to bundling them together under the general description of writing.

Someone who struggles with grammar may possess a remarkable talent for storytelling, while someone capable of producing beautiful prose may have comparatively little ability to construct a compelling narrative. Neither has everything necessary to produce a successful novel independently, but their deficiencies are hardly equivalent. A finished manuscript requires a combination of abilities that seldom receive much attention until someone lacks one of them. A writer needs to understand characters, construct events that follow from believable circumstances, and give readers some reason to care about what happens. They also need to express those elements through language, with enough control over grammar and sentence construction that the reader can follow the story without constantly stumbling over the prose.

Publishing has traditionally rewarded people capable of bringing these abilities together, whether independently or with professional assistance. A poorly written manuscript may never receive serious consideration, even when the underlying story has considerable potential. An elegantly written manuscript with weak characters or an incoherent plot may fare no better, although its technical competence can make those shortcomings less immediately apparent.

Generative AI introduces an intriguing complication. Someone with considerable storytelling ability can now receive assistance with the parts of writing they struggle to perform. The same technology can also help someone with little storytelling talent invent characters, develop conflicts, and construct a novel they might never have conceived independently. Both can produce convincing manuscripts, and both can claim to have used AI in the creative process. The finished work may tell us far less about the abilities of the person whose name appears on the cover than we've become accustomed to assuming.

Where storytelling ends and writing begins

At first, the distinction between creative and technical ability seems straightforward. Storytelling involves imagination, characterization, narrative construction, conflict, and the development of ideas. Technical writing involves spelling, grammar, punctuation, consistency, and the mechanics necessary to make a manuscript readable. Someone may understand why a character would betray a friend without knowing how to communicate that betrayal effectively in prose. They may recognize that revealing information too early would undermine the next chapter, yet struggle to construct a scene that maintains the intended tension.

Those are different kinds of expertise, and an author's proficiency in one doesn't necessarily establish proficiency in the other. The complication arises when we examine the language itself, because prose composition involves considerably more than arranging grammatically correct sentences. Imagine a scene in which a woman returns to her childhood home after twenty years and discovers that it has been demolished. The author knows the scene should communicate loss, although the character is unwilling to acknowledge that loss openly. They understand her history, what brought her back, and how the discovery should affect her. All the necessary narrative information is present, but the author struggles to make the scene emotionally convincing.

A capable prose writer might decide that the woman should search for a tree that once stood beside the driveway. She expected the house to be gone, but finding that the tree has disappeared as well affects her in a way she hadn't anticipated. The character never needs to explain her grief. Her attention to something seemingly ordinary allows the reader to understand what the loss means to her. The tree is a small detail, yet it changes how the scene works. It may even acquire significance elsewhere in the novel, becoming part of the character's relationship with her childhood or the people she left behind. The person who introduced it contributed something meaningful to the story, regardless of whether they were the original author, an editor, or another collaborator.

If the author had already imagined the tree and understood its significance, a collaborator who developed the prose would be expressing an existing creative intention. If the collaborator invented the detail and recognized what it could accomplish, they performed a different kind of creative work. Both arrangements could produce an identical passage, with the same emotional effect on the reader. This is where the tidy division between creative and technical writing begins to fall apart. Literary expression can create meaning that did not exist in the original conception. A writer searching for the right sentence may discover something unexpected about a character, while a conversation written for one purpose may reveal a conflict the author hadn't anticipated. Those discoveries can change the direction of the story, sometimes substantially.

There is a tendency to describe AI prose generation as an advanced form of spellchecking, particularly when defending its use in fiction. I think that comparison becomes untenable once the system begins making decisions about imagery, dialogue, characterization, or the emotional presentation of a scene. Correcting punctuation and composing literary expression involve different levels of creative participation. The fact that both may improve the manuscript doesn't make them equivalent. Still, the difficulty of separating story from prose doesn't establish that a person must excel at both to possess meaningful storytelling ability. It tells us that a novel involves several forms of creativity, and that we need to be more precise about who contributes what when those abilities are distributed among different people or systems.

The author and the editor

Raymond Carver's relationship with Gordon Lish provides a particularly revealing example of how complicated creative attribution can become. Carver's short fiction became known for its economical prose, restrained characterization, and the considerable emotional weight carried by seemingly ordinary events. Lish, his editor, played a substantial role in shaping some of the stories that established that reputation. Their surviving manuscripts and correspondence reveal extensive cuts, altered endings, and revisions that changed how readers encountered the characters and understood their circumstances.

The story Carver submitted as Beginners became What We Talk About When We Talk About Love after Lish's editing. A detailed examination of their manuscripts and correspondence in The New Yorker documents how extensively Lish intervened, removing substantial portions of the text and changing the literary effect of the stories. Carver eventually became sufficiently distressed by the revisions to ask Lish to stop publication of the collection. The disagreement was more complicated than an editor simply imposing his preferences on an unwilling author. Carver had previously valued Lish's judgment and accepted considerable editorial intervention. Their relationship deteriorated as the extent and consequences of that intervention became more difficult for Carver to accept. The surviving correspondence reveals both collaboration and serious disagreement about what the stories should become.

Critics have also disagreed over which versions are stronger. Some prefer the earlier manuscripts, finding greater emotional depth and development in the material Lish removed. Others regard the published versions as more effective because of their restraint. There is no need to settle that dispute to recognize that the changes were consequential. Carver remained the credited author, even though Lish helped shape literary qualities for which the published stories became known. The language on the page was, in places, the result of substantial intervention by someone whose name did not appear as the author.

I'm reluctant to turn this into an argument that Lish should have received coauthorship credit. That would require a different examination of their relationship, agreements, and contributions. The example is useful because it demonstrates how much creative influence can exist within a role we casually describe as editing. A person can alter the artistic identity of a story while remaining formally subordinate to its credited author.

Ghostwriting makes the division of labor more explicit. A person may commission a book, supply its premise or direction, and remain involved throughout development while a professional writer constructs the manuscript. Depending on the arrangement, that writer might do considerably more than translate an existing story into polished language. They may invent scenes, develop characters, resolve structural problems, and introduce ideas that become central to the finished work. The Authors Guild maintains a marketplace for professional ghostwriters, reflecting how established such collaboration has become. That doesn't mean every arrangement is fair, particularly when the person performing substantial creative labor receives little public recognition. It does demonstrate that publishing has long accepted books whose credited authors did not personally perform every significant part of the writing.

An acknowledgment, contractual credit, and actual creative contribution can describe different aspects of the same book. Readers seldom have enough information to reconstruct that history, nor are they necessarily interested in doing so. Most encounter the finished manuscript, the credited author, and whatever account of the book's creation the publisher chooses to provide. Generative AI enters an industry that has already learned to accommodate these ambiguities. What changes is the accessibility of the assistance, its capacity to generate creative material, and the absence of another identifiable human contributor whose role might otherwise be acknowledged.

Assistance, contribution, and control

Suppose a novelist writes a complete manuscript and uses AI to correct grammar, punctuation, and obvious inconsistencies. The system may modify thousands of words without materially changing the characters, plot, or author's creative intentions. The novelist still originated the story, developed its structure, and composed the substantive prose. Their technical deficiencies have been corrected, although the creative work remains substantially their own.

Now consider an author who uses AI only once. They have reached the final chapter and cannot decide how the novel should end. The system proposes a revelation that transforms the meaning of everything that came before it. The author recognizes that the suggestion works, writes the ending themselves, and publishes the novel. The second author may have accepted a far more consequential creative contribution despite having considerably less AI-generated language in the finished manuscript. A single idea can change an entire story, whereas thousands of grammatical corrections may leave its creative substance largely untouched.

This makes the amount of AI-generated or modified prose a questionable proxy for creative contribution. The nature of the work performed is more revealing than the volume of textual changes. There is another distinction hiding inside the idea of control. Someone can possess complete authority over a manuscript while having originated relatively little of its content. A person who asks AI to generate ten endings, selects one, and approves the final result has exercised judgment, but the selection itself doesn't establish that they conceived the ending. The process may involve considerable taste and understanding, particularly if the person can explain why one possibility works and the others fail. It remains a different accomplishment from developing the ending independently.

Now imagine an author who has spent months working through the ending of a complicated novel. They know what each character understands, which conflicts remain unresolved, and what the final scene must communicate. They provide detailed instructions to AI, examine the resulting prose, reject passages that contradict their intentions, and continue revising until the scene behaves as intended. The system is participating in literary expression, sometimes making choices the author hasn't explicitly specified. The author, however, has supplied considerable narrative direction and exercised informed judgment throughout the process. Describing the result merely as AI-generated would conceal much of what the person contributed.

I think this distinction becomes especially important when we consider what it means to approve something. Final authority establishes who gets to decide what appears in the book. It doesn't necessarily tell us who originated the ideas or language. An author might approve an excellent passage without understanding why it works, while another may recognize subtle problems in a passage they lack the technical ability to rewrite independently.

The ability to recognize and direct good writing is a creative skill, albeit one that differs from producing the original material. An experienced developmental editor may possess extraordinary literary judgment while having no interest in writing a novel of their own. A storyteller might understand the precise emotional consequence they want from a scene while struggling to translate that understanding into effective sentences. I don't see a compelling reason to deny those abilities merely because they don't arrive together in the same person. I also wouldn't attribute every accomplishment of an assistant to the person directing it. Both positions would obscure the work that was performed.

A reversal makes the problem clearer. Imagine someone asking AI to invent a novel's premise, characters, conflicts, structure, and ending. They then write the entire manuscript independently, choosing every sentence and composing all the dialogue and descriptions. The prose is human-written, but substantial portions of the underlying narrative originated with AI. Another person might independently develop a story in considerable detail, including its characters, scenes, and intended emotional effects, while relying on AI to produce much of the final prose. Both have made meaningful creative contributions, although in different parts of the process. A classification based entirely on who produced the sentences would favor one kind of human achievement while overlooking another.

This doesn't mean literary expression should receive less recognition than narrative conception. Some novels derive much of their artistic identity from the language itself, and a person who composes that language may contribute far more than the person who supplied a general premise. We need enough information about the actual creative process to understand the difference.

What does human authorship certify?

The Authors Guild's Human Authored certification offers an interesting example of how difficult these distinctions become when an organization must establish practical rules. In May 2026, the Guild clarified its definition to require text authored by human beings, allowing only minimal AI-generated or AI-modified material, such as routine spelling and grammar corrections. The Guild also acknowledged concerns about AI brainstorming and outlining, recognizing that those activities could become extensive enough to replace substantial human creativity even when the final text was written by a person.

The clarification removed explicit references to brainstorming and outlining from the certification definition. This avoided granting those activities blanket approval, although the operative standard still centered on human-composed text. I find that acknowledgment revealing. The Guild recognizes that AI can make important creative contributions before a manuscript exists. Yet determining whether those contributions displaced meaningful human creativity is considerably harder than asking whether the final sentences were generated by a machine.

Amazon takes a different approach through its Kindle Direct Publishing guidelines. Amazon distinguishes AI-generated content from AI-assisted work. Publishers must disclose generated content to Amazon, including material subsequently subjected to substantial human editing, while AI used to refine or correct content originally created by the author falls into the assistance category.

These policies serve different purposes. A voluntary certification can help readers identify books meeting a human-composition standard. A publishing platform needs rules that can be explained to users and applied across a large marketplace. Neither organization is obliged to resolve every philosophical disagreement about creativity before establishing a workable policy. Problems arise when those categories are used to make broader claims than they can support.

A ghostwritten book may contain entirely human-composed prose even when the credited author performed relatively little of the writing. A novelist who personally developed and drafted an entire manuscript might fail a human-written certification after accepting extensive AI prose revisions. Both classifications could be correct under the applicable standard, while providing an incomplete description of the credited person's contribution. Readers may reasonably prefer books composed entirely by humans. Someone who admires an author's particular use of language may want assurance that those sentences represent the author's own literary work. Others may be more interested in the narrative, its characters, or the ideas developed through the story. The preference for human composition doesn't require a belief that every AI-assisted novel must be inferior.

There are also legitimate concerns about the materials used to train generative models, the rights of the people whose work contributed to that training, and the economic consequences for professional writers and editors. Those issues deserve serious examination, although they are separate from whether a particular finished novel is compelling. A book can raise questions about its production while still succeeding artistically. What I find difficult to justify is treating the presence of AI assistance as a complete account of the human creativity involved. The work performed by the person, the work performed by the system, and the way the resulting book is represented all need to be considered.

The distinction becomes more complicated when publishers or competitions attempt to verify authorship through AI detection. Even an accurate classification of generated prose tells us relatively little about who conceived the story, developed its characters, or directed the creative process. Someone could construct an entire novel from AI-generated narrative ideas while personally composing every sentence. Another could originate the complete story and rely on AI to help express it. A detector might identify generated language in the second manuscript while finding little to question in the first, despite the very different creative contributions involved.

I've examined the reliability and consequences of AI detection in AI Detectors Are Bullshit, including the Commonwealth Short Story Prize controversy and the evidence that contradicted accusations made by Pangram. That investigation raises serious questions about treating detector results as proof of misconduct. Here, the limitation is more fundamental. Even if a detector correctly identifies AI-generated writing, it cannot establish whether the author's involvement amounted to technical assistance, substantial creative direction, or little more than approving a finished manuscript. Those distinctions depend on how the work was created, information the finished prose cannot reliably provide.

When polished prose becomes inexpensive

There is a broader consequence to making competent prose accessible to almost anyone with a computer. A polished manuscript once established that someone involved in its production possessed the necessary writing skills, or had obtained professional assistance from someone who did. That person might have been the credited author, an editor, or a ghostwriter. The manuscript itself never guaranteed that the credited author independently possessed every ability reflected in the finished text, although the cost and effort of producing such work imposed practical limits.

Generative AI reduces those limits. A person who struggles with grammar can obtain corrections immediately. Someone who has difficulty constructing dialogue can request alternatives. A writer who cannot resolve a plot problem can ask for suggestions, evaluate the results, and incorporate whichever ideas they find convincing. Someone with little creative vision can use the same tools to generate much of a novel from a relatively simple premise.

The consequences are unlikely to be entirely beneficial or entirely destructive. Increasing access to literary production may allow stories to reach readers that would otherwise have remained unfinished. It can also increase the production of technically competent manuscripts whose creators contributed relatively little original storytelling.

There is research suggesting another complication. In a 2024 study published in Science Advances, Anil Doshi and Oliver Hauser examined how access to AI-generated ideas influenced short-story writing. Participants with access to those ideas produced stories that received higher average evaluations for creativity, writing quality, and enjoyment. The improvements were particularly pronounced among participants who initially scored lower on creativity. The researchers also found that AI-assisted stories became more similar to one another.

That combination is worth considering. A tool can help individuals improve their work while encouraging different people toward similar creative choices. The study involved short fiction produced under experimental conditions, so it would be premature to assume the same result applies to novels developed over years. It does, however, demonstrate that improved individual results and increased collective originality aren't necessarily the same outcome.

An author may encounter a related problem during revision. A system can make awkward prose more fluent while removing something distinctive about the narrator. It may replace uncertainty with explanation or reshape an uncomfortable ending into something more conventional. The result may read better according to ordinary standards while serving the author's creative intentions less effectively.

A person who understands those intentions can recognize the damage and reject the revision. Someone who accepts whatever sounds polished may gradually allow the system to determine aspects of the work they believed they still controlled. This suggests that creative judgment becomes increasingly important as the mechanical production of competent language becomes easier. An author who relies on assistance may need considerable knowledge of storytelling and literary expression to recognize when the assistance is producing the wrong result. They don't necessarily need to execute every correction personally, but they do need to understand what they are accepting.

The author behind the manuscript

Return to the aspiring novelist who has spent years developing a story but struggles to express it in polished prose. Perhaps AI helps correct their grammar, restructures awkward sentences, or offers alternatives that communicate their intentions more effectively. They may reject suggestions, rewrite passages, and spend considerable time directing revisions. Some of those revisions may introduce creative material the author never originally conceived, while others may simply make the existing story easier to understand.

I see no reason to dismiss the person's storytelling ability because they required assistance with literary execution. At the same time, it would be misleading to attribute every creative accomplishment in the finished manuscript to them if another contributor, human or artificial, supplied substantial parts of the expression or narrative.

The distinction becomes especially interesting when we compare that person with someone who has little story to tell but can produce flawless prose. The second person may possess a kind of literary expertise the first lacks, and there are novels in which that expertise is central to the artistic achievement. There are also stories whose power comes principally from their characters, conflicts, and the consequences of their events. Neither form of accomplishment needs to invalidate the other.

Publishing has accommodated uneven distributions of creative ability for a long time. Authors have dictated works, collaborated with editors, commissioned ghostwriters, and relied on other people to strengthen weaknesses in their manuscripts. Those arrangements differ in important ethical and creative respects, yet they all complicate the assumption that the name on the cover tells us exactly who performed the work.

AI changes the economics and accessibility of that assistance. It also makes the division of creative labor less visible, because an individual can now perform an entire collaborative process privately with software. A manuscript may arrive with excellent grammar, convincing dialogue, and carefully constructed scenes without establishing that the person submitting it possesses those abilities independently. That person may nevertheless have conceived an exceptional story and exercised considerable judgment in bringing it to publication. Another may have delegated nearly all the creative work while retaining complete control over the finished product.

The polished manuscript can no longer serve as reliable evidence of which situation we're encountering, although the relationship between prose proficiency and credited authorship was never as straightforward as we sometimes assumed. The name on the cover remains, but what it tells us about the person behind the book has become considerably harder to determine.