When a novelist asks an AI chatbot to generate a full-length work, the expectation is often that the machine will do the heavy lifting: the sentences, the scenes, the characters, and the hard-to-engineer narrative arc. But one writer’s experiment—described in a Mother Jones report—suggests that the “machine” may not be the limiting factor so much as the scaffolding around it. The prose may arrive quickly, yet turning that output into a convincing novel still depends on careful judgment about length, pacing, dialogue, and continuity.
The writer’s starting point was modest: the AI produced a manuscript that clocked in at roughly 6,000 words. That length, the writer notes, was “much too short” to qualify as a novel by typical expectations. The chatbot agreed, reframing the output as closer to a novella and suggesting that a “proper novel” would run somewhere around 70,000 to 90,000 words. In other words, the first fault line in the experiment wasn’t grammar or style—it was scale, the fundamental question of what kind of story form the draft was actually supporting. News Source
From there, the writer asked the AI to expand the draft into something closer to a conventional book: full-length scenes, developed secondary characters, and “granular detail” that gives fiction the sense of having been lived in. The experiment reads like a conversation staged for creativity—at least on the surface. The writer reports responding as if to a person rather than a tool, prompting the system to build out “real scenes” and broaden the cast. But as the draft grew, structural questions remained central: more words didn’t automatically translate into richer fiction.
Even with expansion, the writer identified “obvious flaws.” Scenes were not fully realized, with action compressed so tightly that it didn’t breathe. Chapters were described as too short, and dialogue—an essential mechanism for character voice and relationship dynamics—was insufficient. Additional room for the secondary characters also seemed necessary. Particular word-choice issues surfaced too, including a repeated use of the term “particular,” which can make a draft feel formulaic when it shows up too often without emotional justification or variety. News Source
Continuity, too, proved vulnerable. A few chronological inconsistencies pulled the writer out of the narrative’s internal logic. Near the end, the writer found the story’s final stretch confusing in spots. The implication is stark: even if AI can produce text rapidly, it can still falter when the work requires long-range coordination—keeping timelines straight, sustaining escalation, and making sure the ending resolves in a way that feels inevitable rather than abruptly assembled.
The daily rhythm of the experiment is part of what makes it illuminating. The writer continued the routine of working each morning for several hours writing—then pivoted in the afternoon to using AI to generate variations. The chatbot, in this workflow, wasn’t treated as a plug-and-play author. Instead, it produced multiple drafts of the same kind of content: alternative versions of scenes and even competing versions of conversations. The writer also asked questions about structure, soliciting options rather than accepting a single “solution.” In that sense, the AI functioned as an idea generator and storyboard engine, offering directions that the human writer could choose to follow or ignore. News Source
Crucially, the writer says the output was not pasted wholesale into the draft. Instead, the AI’s value lay in mapping possibilities: different structural decisions, emotional interpretations, and conceptual insights. The prose, the writer reports, remained “always” their own, even as AI provided a kind of parallel design process. That distinction—AI contributing to the plan and the human writing the sentences—frames the experiment as less about replacement and more about augmentation, with the human responsible for final coherence.
Yet the report also gestures toward why those separations matter. If the AI generates a passage that isn’t fully realized at the scene level, the human editor has to decide whether the problem is local—missing dialogue, insufficient development—or systemic—pacing that never gives characters room to act believably. When chapters are too short and action overly compacted, a writer may need to expand not just content but cause-and-effect relationships: set up stakes, deepen motivations, and provide transition space so readers can track how events change people.
Likewise, the chronological inconsistencies and confusion near the end point to the difficulty of long-form consistency. Novels demand that every new element harmonize with what came before. Even a minor timeline error can distort character decisions, undermine trust in the narrator’s reliability, or make earlier foreshadowing feel pointless. The writer’s speculation is telling: had the AI been the one to fully expand the draft, some errors might have been remedied. But the experiment’s central lesson is that expansion alone is not a safeguard; the craft of editing, rebalancing, and verifying continuity remains decisive.
Ultimately, the writer’s conclusion—explicitly or implicitly—reads as a nuanced assessment of AI creativity. The chatbot could help generate scene frameworks, suggest alternate dialogue, and accelerate ideation. But the experiment underscores a boundary: a novel is not merely a large amount of text. It is a carefully governed system of length, rhythm, dialogue density, character work, and chronological stability. In that system, the “big problem” is less about whether AI can write and more about whether the draft can be made to behave like the kind of story readers expect from a novel.
SHOP AMAZON BEST SELLERS, CLICK TO BUY FROM AMAZON.
SHOP AMAZON BEST SELLERS, CLICK TO BUY FROM AMAZON.
