AI应用 · 生产力

Notion AI每天生成600万篇文档
85%用户已离不开

Notion AI Generates 6M Docs Daily
85% of Users Can't Live Without It

每天600万篇文档由AI协助生成,活跃用户85%在用AI写作。不是AI去找人,是AI长在光标里,等着你用。这是AI应用从工具变成基础设施的分界。

6 million documents generated with AI assistance daily, 85% of active users use AI writing. It's not people going to AI — it's AI living in the cursor, waiting for you to use it. This is the dividing line where AI apps go from tools to infrastructure.

No.047 2026.09.07 约 5 分钟阅读 ~5 min read

有多少人的写作,正在偷偷依赖 Notion AI?答案可能比你想的多得多。

9月6日,Notion 官方披露了一组数据,有点惊人:每天有 600 万篇文档由 Notion AI 协助生成,活跃用户中的 AI 功能采用率达到了 85%。换句话说,十个用 Notion 的人里,有八个半在用 AI 写东西。

这不是"偶尔用一下试试看"的那种试用。这是"每天写文档就开着 AI"的深度使用。Notion AI 从一个"看起来不错"的附加功能,已经变成了用户离不开的基础设施。

更有意思的是它背后的产品逻辑:AI 不是一个单独的功能按钮,是长在文字光标里的。你不需要切换到另一个窗口、不需要复制粘贴、不需要专门调用。你在写文档的任何地方,按一个快捷键,AI 就出现在光标后面,接着你的思路往下写。

为什么是 Notion 先跑出来?

AI 写作工具那么多,为什么 Notion AI 的渗透率这么高?核心原因就一个:它不打扰你。

市面上大多数 AI 写作工具的产品形态是"你给我个主题,我给你生成一篇文章"——你得主动去找它、得切换上下文、得把生成的内容粘回来。这中间的摩擦,哪怕只有几秒钟,也足以让大多数人在"忙的时候"选择不用。

Notion AI 不一样。它就在你的文档里、就在你的光标旁边。你写着写着卡住了,按一下 Tab,它接着你的上下文继续写。你觉得某段话写得不好,选中按个快捷键,它帮你重写。整个过程你不用离开文档、不用切换窗口、不用复制粘贴。AI 不是一个工具,是你写作流的延伸。

这个产品设计思路看似简单,实际上是 AI 应用的一个核心分界:是让人去找 AI,还是让 AI 去找人?前者是工具,后者是基础设施。工具需要你主动打开,基础设施是你用的时候它就在那里。

85% 渗透率意味着什么

85% 这个数字,在 AI 应用领域是非常罕见的。大多数 SaaS 产品加了 AI 功能之后,实际使用率可能连 10% 都不到。用户觉得"挺有意思",然后就忘了。

Notion 能做到 85%,说明三件事:

第一,AI 写作已经从"锦上添花"变成了"刚需"。当十个用户里有八个半天天用,它就不再是一个可选项了。以后用户选写作工具,"有没有好用的 AI 功能"会是一个决定性因素,不是加分项。

第二,"上下文原生"是 AI 应用的正确打开方式。Notion AI 之所以好用,不是因为它的模型比别人强多少,而是因为它就在你的文档里、知道你在写什么、知道上下文。同样的模型,放在独立的聊天机器人里和嵌在文档里,用户体验天差地别。

第三,办公软件的 AI 竞赛,已经从"有没有"进入了"好不好用"的阶段。去年大家还在比"谁家先上线 AI 功能",今年比的是"谁家的 AI 真的能帮上忙、真的有人用"。Notion 用 85% 的渗透率给自己打了个分,这个分数让其他竞品压力很大。

下一站:从"帮你写"到"替你想"

Notion AI 现在还处在"帮你写"的阶段——你有思路,它帮你落地、帮你润色、帮你扩展。但下一阶段,AI 会往"替你想"的方向走。

想象一下:你在写周报,AI 自动把你这周做过的项目、开过的会、聊过的天全部整理好,生成一份初稿;你在写方案,AI 自动调取你知识库中相关的所有文档,帮你补全背景和数据;你在写邮件,AI 自动根据对方的身份和历史对话,调整语气和措辞。

这些事情,Notion 天然就有数据基础——你的笔记、文档、项目、知识库全在它里面。当 AI 不只是"会写字",而是"懂你的工作",那办公软件的定义就彻底变了。

当然,这条路也有隐忧。当 85% 的文档都有 AI 参与,那"人的写作能力"会不会退化?当 AI 帮你想了,你自己的思考会不会变懒?这些问题没有标准答案。但有一件事是确定的:回不去了。一旦你试过了"光标后面跟着一个 AI"的写作体验,再回到纯手动写作,就像用惯了智能手机再回去用功能机一样。

明天见。

How many people's writing is secretly relying on Notion AI? The answer might be higher than you think.

On September 6, Notion officially disclosed some numbers that are kind of astonishing: 6 million documents per day are generated with Notion AI assistance, and AI feature adoption among active users has reached 85%. In other words, out of ten Notion users, eight and a half are using AI to write things.

This isn't the "let me try it out once" kind of trial usage. This is "I turn on AI every time I write a document" deep usage. Notion AI has gone from a "nice-to-have" add-on to infrastructure users can't live without.

What's even more interesting is the product logic behind it: AI isn't a separate feature button — it lives right in the text cursor. You don't need to switch to another window, copy-paste, or specially invoke it. Anywhere you're writing in a document, hit a keyboard shortcut, and AI appears right after your cursor, continuing your train of thought.

Why Did Notion Break Out First?

There are so many AI writing tools out there. Why is Notion AI's penetration so high? The core reason is one thing: it doesn't interrupt you.

Most AI writing tools on the market follow the product form of "you give me a topic, I generate an article for you" — you have to actively go find it, switch context, paste the generated content back. That friction in between, even if it's just a few seconds, is enough to make most people choose not to use it when they're busy.

Notion AI is different. It's right there in your document, right next to your cursor. You're writing and you get stuck, hit Tab, and it continues where your context left off. You think a paragraph is poorly written, select it and hit a keyboard shortcut, and it rewrites it for you. The whole process you never leave the document, never switch windows, never copy-paste. AI isn't a tool — it's an extension of your writing flow.

This product design approach seems simple, but it's actually a core dividing line in AI applications: do you make people go find AI, or do you make AI come find people? The former is a tool. The latter is infrastructure. Tools require you to actively open them; infrastructure is just there when you need it.

What Does 85% Penetration Mean?

The 85% number is extremely rare in the AI application space. Most SaaS products, after adding AI features, might see actual usage rates below 10%. Users think "that's kinda cool," then forget about it.

Notion reaching 85% means three things:

First, AI writing has gone from "nice to have" to "must-have." When eight and a half out of ten users use it every day, it's no longer optional. Going forward, when users choose a writing tool, "does it have good AI features" will be a deciding factor, not a bonus.

Second, "context-native" is the right way to do AI applications. Notion AI is good not because its model is that much better than others, but because it's inside your documents, knows what you're writing about, knows the context. The same model, placed in a standalone chatbot versus embedded in documents, delivers a night-and-day user experience.

Third, the AI race in office software has moved from "do you have it?" to "is it actually good?" Last year everyone was competing on "who launches AI features first." This year it's about "whose AI actually helps and actually gets used." Notion scored itself an 85% penetration rate, and that score puts a lot of pressure on competitors.

Next Stop: From "Helping You Write" to "Thinking for You"

Notion AI is still in the "helping you write" phase — you have the idea, it helps you execute, polish, expand. But the next phase, AI will move toward "thinking for you."

Imagine: you're writing your weekly report, and AI automatically compiles all the projects you worked on this week, meetings you had, conversations you had, and generates a first draft. You're writing a proposal, and AI automatically pulls all relevant documents from your knowledge base to help fill in background and data. You're writing an email, and AI automatically adjusts tone and wording based on who you're writing to and your conversation history.

These are things Notion naturally has the data foundation for — your notes, documents, projects, knowledge base are all inside it. When AI isn't just "good at writing" but "understands your work," that's when the definition of office software completely changes.

Of course, there are concerns along this path too. When 85% of documents have AI participation, will "human writing ability" degrade? When AI thinks for you, will your own thinking get lazier? There are no easy answers to these questions. But one thing is certain: there's no going back. Once you've experienced writing with an AI following your cursor, going back to pure manual writing feels like going from a smartphone back to a feature phone.

See you tomorrow.

AI应用的核心分界:是让人去找AI,还是让AI去找人?前者是工具,后者是基础设施。

—— Dawn Vision编辑部

The core dividing line of AI apps: do people go find AI, or does AI come find people? The former is a tool, the latter is infrastructure.

— The Dawn Vision Editorial Desk
Notion AI · 85%渗透率 · 600万文档 · AI写作 · 上下文原生 · 光标AI · 办公软件 · 基础设施化 · 从帮你写到替你想
Notion AI · 85% penetration · 6M documents · AI writing · context-native · cursor AI · office software · infrastructure · from helping write to thinking for you
Sources · 信源 Sources

本文基于 Dawn Vision 认知引擎处理的 8 个源信号生成,经编辑部人工审核。素材来源:Notion官方、The Verge、TechCrunch。

This article was generated by the Dawn Vision cognitive engine processing 8 source signals, with human editorial review. Sources: Notion Official, The Verge, TechCrunch.