一套针对优秀文章的深度追问提示词组合：依次获取行动建议、提炼反常识与高信息熵内容、追问案例论据、生成并引用原文回答问题，全部用Markdown结构化输出，适合深度阅读与知识挖掘。

中文版提示词

以下是基于一篇文章进行深度追问的系列提示词，可依次使用：

1. 生成行动建议：
请一步一步仔细思考，基于文章观点给我几个行动建议，请不要遗漏任何一条，不要担心回复太长，回复要求有逻辑、结构化，用Markdown展示。

2. 提炼反常识与高信息熵内容：
请一步一步仔细思考，提炼文章中的反常识观点和高信息熵内容，请不要遗漏任何一条，先给结论，再给论据，尽量在原文基础上提炼，不要省略数字和年份等关键信息，不要担心回复太长，回复要求有逻辑、结构化，用Markdown展示。

3. 追问案例与论据：
以上信息的案例和论据都是什么？请一步一步仔细思考回答，引用原文案例和论据，每个案例和论据都用一句话总结，直到找出文章中所有案例和论据。数字和年份都是关键信息，总结时请不要概括，直接引用。不要担心回复太长，回复要求有逻辑、结构化，可以用Markdown展示。

4. 生成问题并引用原文回答：
你还可以让AI基于文章内容提出几个问题，再让AI引用原文来回答这些问题。

英文版提示词

Here is a set of prompts for deeply interrogating an article; use them in order:

1. Generate action suggestions:
Think carefully step by step and give me several action suggestions based on the article's viewpoints. Do not omit any, do not worry about the reply being too long, and make the reply logical, structured, and formatted in Markdown.

2. Extract counterintuitive viewpoints and high-information-density content:
Think carefully step by step and extract the counterintuitive viewpoints and high-information-density content from the article. Do not omit any; give the conclusion first, then the evidence. Extract as much as possible from the original text, and do not omit key information such as numbers and years. Do not worry about the reply being too long; make it logical, structured, and formatted in Markdown.

3. Follow up on examples and evidence:
What are the examples and evidence for the above information? Think carefully step by step and answer by citing the original examples and evidence, summarizing each one in a single sentence until you have found all examples and evidence in the article. Numbers and years are key information; do not paraphrase them, quote them directly. Do not worry about the reply being too long; make it logical, structured, and formatted in Markdown.

4. Generate questions and answer with citations:
You can also ask the AI to raise several questions based on the article, then have the AI answer those questions by citing the original text.

🛠️ **适用 AI 工具**：ChatGPT、Claude、Kimi、DeepSeek、通义千问、文心一言