AI-Assisted Nonfiction Material Mining
What it is
AI-assisted nonfiction material mining is using AI to help collect, clean, search, translate, and interrogate source material for long-form nonfiction while leaving the argument, scene selection, voice, and final judgment with the human writer.
Why it matters
The EP18 case is valuable because it is not an "AI wrote the article" story. The author says the final nonfiction piece did not contain AI-written prose, but AI still made the project possible by helping process a large, multilingual source base and turn scattered material into usable writing paths.
This is a better model for Jean's writing workflow than generic drafting: AI helps create a richer block of material, then the writer sculpts it.
Evidence across sources
| Source | Evidence | What it supports |
|---|---|---|
| EP18 与 AI 一起写非虚构 | The writer used AI for transcript cleanup, course framing, multilingual keyword generation, source search, and material excavation, while retaining scene judgment and prose ownership. | AI can be an upstream research and material-mining partner without becoming the author. |
Process pattern
- 现场采集：录音、拍照、笔记、访谈和实地观察仍是 nonfiction 的地基。
- 资料清洗：AI 可帮助修转写、整理字段、建立章节或时间线。
- 关键词扩展：AI 生成多语言关键词和历史/专业术语，扩大可搜索材料范围。
- 材料挖掘：AI 在大量资料中找原文、细节、反常点和可能的故事线索。
- 结构试探：AI 可帮助尝试课程框架或文章结构，但不决定最终主题。
- 人类取舍：画面感、细节是否有意思、文章钥匙、朗读后的节奏判断仍由作者承担。
Boundaries
- AI does not replace fieldwork.
- AI does not own the sentence-level final voice.
- AI can expand the search space, but the writer decides what matters.
- Historical or factual claims still need source verification before publication.
Open questions
- How should Jean's output pipeline preserve raw material, extracted fragments, and final draft separately?
- Which parts of this process should become a reusable
writing-shapeoroutput-draftchecklist? - How much AI-generated structure is helpful before it starts flattening the writer's original question?