
来自 @Ihtesham 分享的高质量提示词,青小蛙觉得写的非常赞啊,值得仔细学习一下。
重新整理过了。

注意,以下提示词虽然也适合于其他 AI 工具,但 NotebookLM 的与众不同之处在于:
你给它什么资料(文档、网页、视频、图片…),它就只在这些资料里思考,不会自由发挥,瞎编乱造。
让 AI 扮演某个领域 15 年经验的专家,从多份资料中提炼出:
👉 适合:
快速判断一堆材料「有没有真货」、值不值得继续深挖。
“You are a [field] expert with 15 years of experience. Analyze these sources and identify the 3 core insights that practitioners in this field would immediately recognize as groundbreaking. For each insight, explain why it matters and what conventional wisdom it challenges.”
“您是一位拥有15年经验的[领域]专家。请分析这些资料,并找出该领域从业者一眼就能看出具有突破性意义的3个核心见解。对于每个见解,请解释其重要性以及它挑战了哪些传统观念。”
要求 AI:
👉 适合:
综述论文、政策分析、行业报告、技术路线之争。
Compare these sources and identify every point where they contradict each other. For each contradiction, explain which source has stronger evidence and why. If both are credible, explain what factors might explain the disagreement.
“比较这些资料,找出它们之间所有相互矛盾的地方。对于每一个矛盾之处,解释哪个资料的证据更有力,以及原因。如果两者都可信,解释可能导致分歧的因素。”
从所有资料中提取:
并整理成一步步的实施计划。
👉 适合:
“看懂了,但不知道怎么落地”的所有场景。
Extract every actionable step, tool, framework, and technique mentioned across all sources. Organize them into a step-by-step implementation plan with prerequisites, expected outcomes, and potential pitfalls for each step.
提取所有来源中提到的每一个可操作的步骤、工具、框架和技术。将它们整理成一个循序渐进的实施计划,每个步骤都应包含前提条件、预期结果和潜在风险。
让 AI 生成:
并优先挑选:
👉 适合:
写论文选题、创业方向、产品差异化、深度内容选角度。
Based on these sources, generate 15 questions that an expert would ask but that these sources DON’T answer. Prioritize questions that would advance the field or reveal critical gaps in current understanding.
基于这些资料,提出15个专家会问但这些资料没有回答的问题。优先考虑那些能够推动该领域发展或揭示当前理解中关键空白的问题。
要求 AI:
👉 适合:
识别系统性风险、认知偏差、共识幻觉。
Identify every unstated assumption in these sources. For each assumption, rate how critical it is (1-10) and how likely it is to be wrong. Explain what would change if that assumption were false.
找出这些资料中所有未明确说明的假设。对于每个假设,评估其重要性(1-10)以及其错误的可能性。解释如果该假设为假,将会发生什么变化。
让 AI 构建一个完整框架,包含:
👉 适合:
把零散观点变成「能反复使用的认知工具」。
Create a comprehensive framework that integrates all concepts from these sources. Include: key components, relationships between components, decision trees for application, and edge cases where the framework breaks down.
创建一个综合框架,整合来自这些来源的所有概念。包括:关键组成部分、组成部分之间的关系、应用决策树以及框架失效的极端情况down.
对每一个主要结论:
👉 适合:
反营销、反“专家胡说”、防被带节奏。
For every major claim in these sources, extract the supporting evidence and rate its strength (anecdotal, correlational, experimental, meta-analysis). Flag any claims with weak evidence that are stated with high confidence.
针对这些来源中的每一项主要论断,提取其支持证据并评估其强度(轶事证据、相关性证据、实验证据、荟萃分析证据)。标记出任何证据薄弱但置信度很高的论断。
把同一批结论分别翻译给:
每一版都只讲:
👉 适合:
汇报、产品文档、技术传播、对齐认知。
Translate the insights from these sources for three different audiences: [executives, engineers, end-users]. For each audience, focus on what they specifically care about and use language/examples they’ll immediately understand.
将这些来源的见解转化为三种不同受众群体能够理解的内容:[高管、工程师、最终用户]。针对每种受众群体,重点关注他们具体关心的内容,并使用他们能够立即理解的语言/示例。
从所有资料中,提取一切与时间有关的信息,包括:
把这些信息整理成一条完整的演化时间线不是简单罗列,而是展示:这个领域 / 主题是如何一步步发展到今天的
标出“加速节点”,也就是:
👉 适合解决的问题:
👉 特别适合:
Extract every date, event, milestone, and temporal reference from these sources. Build a comprehensive timeline showing how this field/topic evolved. Identify acceleration points where progress dramatically increased.
从这些资料中提取所有日期、事件、里程碑和时间参考信息。构建一个全面的时间线,展示该领域/主题的发展历程。找出进展显著加速的节点。
它要求 AI 对所有资料逐条做“拆解”,找出:
并且对每一个问题,都要回答一件事:
👉 如果要让这个结论成立,还需要补充什么证据?
这一步的重点不是“否定”,而是:
👉 适合解决的问题:
👉 特别适合:
Act as a harsh peer reviewer. Identify every methodological flaw, logical gap, overclaim, and unsupported leap in these sources. For each weakness, suggest what additional evidence would be needed to strengthen the argument.
扮演严苛的同行评审员角色。指出这些文献中所有的方法论缺陷、逻辑漏洞、夸大其词和缺乏依据的推断。针对每一个缺陷,提出需要哪些补充证据来加强论点。

来自 @Ihtesham 分享的高质量提示词,青小蛙觉得写的非常赞啊,值得仔细学习一下。
重新整理过了。

注意,以下提示词虽然也适合于其他 AI 工具,但 NotebookLM 的与众不同之处在于:
你给它什么资料(文档、网页、视频、图片…),它就只在这些资料里思考,不会自由发挥,瞎编乱造。
让 AI 扮演某个领域 15 年经验的专家,从多份资料中提炼出:
👉 适合:
快速判断一堆材料「有没有真货」、值不值得继续深挖。
“You are a [field] expert with 15 years of experience. Analyze these sources and identify the 3 core insights that practitioners in this field would immediately recognize as groundbreaking. For each insight, explain why it matters and what conventional wisdom it challenges.”
“您是一位拥有15年经验的[领域]专家。请分析这些资料,并找出该领域从业者一眼就能看出具有突破性意义的3个核心见解。对于每个见解,请解释其重要性以及它挑战了哪些传统观念。”
要求 AI:
👉 适合:
综述论文、政策分析、行业报告、技术路线之争。
Compare these sources and identify every point where they contradict each other. For each contradiction, explain which source has stronger evidence and why. If both are credible, explain what factors might explain the disagreement.
“比较这些资料,找出它们之间所有相互矛盾的地方。对于每一个矛盾之处,解释哪个资料的证据更有力,以及原因。如果两者都可信,解释可能导致分歧的因素。”
从所有资料中提取:
并整理成一步步的实施计划。
👉 适合:
“看懂了,但不知道怎么落地”的所有场景。
Extract every actionable step, tool, framework, and technique mentioned across all sources. Organize them into a step-by-step implementation plan with prerequisites, expected outcomes, and potential pitfalls for each step.
提取所有来源中提到的每一个可操作的步骤、工具、框架和技术。将它们整理成一个循序渐进的实施计划,每个步骤都应包含前提条件、预期结果和潜在风险。
让 AI 生成:
并优先挑选:
👉 适合:
写论文选题、创业方向、产品差异化、深度内容选角度。
Based on these sources, generate 15 questions that an expert would ask but that these sources DON’T answer. Prioritize questions that would advance the field or reveal critical gaps in current understanding.
基于这些资料,提出15个专家会问但这些资料没有回答的问题。优先考虑那些能够推动该领域发展或揭示当前理解中关键空白的问题。
要求 AI:
👉 适合:
识别系统性风险、认知偏差、共识幻觉。
Identify every unstated assumption in these sources. For each assumption, rate how critical it is (1-10) and how likely it is to be wrong. Explain what would change if that assumption were false.
找出这些资料中所有未明确说明的假设。对于每个假设,评估其重要性(1-10)以及其错误的可能性。解释如果该假设为假,将会发生什么变化。
让 AI 构建一个完整框架,包含:
👉 适合:
把零散观点变成「能反复使用的认知工具」。
Create a comprehensive framework that integrates all concepts from these sources. Include: key components, relationships between components, decision trees for application, and edge cases where the framework breaks down.
创建一个综合框架,整合来自这些来源的所有概念。包括:关键组成部分、组成部分之间的关系、应用决策树以及框架失效的极端情况down.
对每一个主要结论:
👉 适合:
反营销、反“专家胡说”、防被带节奏。
For every major claim in these sources, extract the supporting evidence and rate its strength (anecdotal, correlational, experimental, meta-analysis). Flag any claims with weak evidence that are stated with high confidence.
针对这些来源中的每一项主要论断,提取其支持证据并评估其强度(轶事证据、相关性证据、实验证据、荟萃分析证据)。标记出任何证据薄弱但置信度很高的论断。
把同一批结论分别翻译给:
每一版都只讲:
👉 适合:
汇报、产品文档、技术传播、对齐认知。
Translate the insights from these sources for three different audiences: [executives, engineers, end-users]. For each audience, focus on what they specifically care about and use language/examples they’ll immediately understand.
将这些来源的见解转化为三种不同受众群体能够理解的内容:[高管、工程师、最终用户]。针对每种受众群体,重点关注他们具体关心的内容,并使用他们能够立即理解的语言/示例。
从所有资料中,提取一切与时间有关的信息,包括:
把这些信息整理成一条完整的演化时间线不是简单罗列,而是展示:这个领域 / 主题是如何一步步发展到今天的
标出“加速节点”,也就是:
👉 适合解决的问题:
👉 特别适合:
Extract every date, event, milestone, and temporal reference from these sources. Build a comprehensive timeline showing how this field/topic evolved. Identify acceleration points where progress dramatically increased.
从这些资料中提取所有日期、事件、里程碑和时间参考信息。构建一个全面的时间线,展示该领域/主题的发展历程。找出进展显著加速的节点。
它要求 AI 对所有资料逐条做“拆解”,找出:
并且对每一个问题,都要回答一件事:
👉 如果要让这个结论成立,还需要补充什么证据?
这一步的重点不是“否定”,而是:
👉 适合解决的问题:
👉 特别适合:
Act as a harsh peer reviewer. Identify every methodological flaw, logical gap, overclaim, and unsupported leap in these sources. For each weakness, suggest what additional evidence would be needed to strengthen the argument.
扮演严苛的同行评审员角色。指出这些文献中所有的方法论缺陷、逻辑漏洞、夸大其词和缺乏依据的推断。针对每一个缺陷,提出需要哪些补充证据来加强论点。