Bing Text CreatorBing Text Creator
When a search engine learns to create, where should a search end?
Making generated content live inside search — and making it worth keeping.
当搜索引擎学会创作，一次搜索的终点应该在哪里？
让生成的内容活在搜索里——并且让它值得被留下。
Supply side — search has to prove it isn't replaceable. In the age of AI, if a chat tool can just answer you, why search at all? "Search + AI" had to create new value — not simply be replaced. 供给侧——搜索得证明自己不可替代。AI 时代，如果一个聊天工具能直接回答你，那为什么还要搜索？ "搜索 + AI" 必须创造新价值，而不只是被替代。
Demand side — users are shifting from finding content to creating it. In the UGC era, what they want is authentic, created things they can use and adapt — not a standard answer. 需求侧——用户正从"搜索内容"转向"创造内容"。在 UGC 时代，他们要的是能拿来用、拿来改的真实创造物， 而不是标准答案。
Both forces point to the same move: AIGC. It turns search from "handing you information" into "helping you make the thing you want." Text Creator is that bet. 两股力量都指向同一个动作：AIGC。让搜索从"把信息给你"，变成"帮你造出你想要的东西"。 Text Creator 就是这个下注。
Bing Text Creator is a Creator Answer that lives on the Bing search results page (SERP). When someone searches with the intent to create something—draft an email, write a toast, outline a weekly plan, tell a bedtime story, etc.—Creator generates, edits, and helps them use that content right inside the results page. Bing Text Creator 是一个落在 Bing 搜索结果页（SERP）上的 Creator Answer：当用户带着创作意图搜索—— 写封邮件、拟段祝福、起草一份周计划、写个睡前故事等等——Creator 直接在结果页里生成、编辑，并帮他把内容用起来。
Everything started with an easily overlooked observation: a Bing user arrives with a query—not, as a ChatGPT user might, with an intent to create. 一切始于一个容易被忽略的观察：Bing 用户是带着一个 query 走进来的—— 而不是像 ChatGPT 用户那样，可能天然带着创作意图。
That difference changes everything. A blank input box is freedom for a ChatGPT user, but a barrier for a search user: they don't even know creation is possible here, let alone what they could create. So Creator could never be about "making the generation box bigger." It had to answer a different question: how do you lower the barrier to creation in a search context—and make that creation genuinely matter to search? 这个差别改变了一切。空白输入框对 ChatGPT 用户是自由，对搜索用户却是门槛：他们甚至不知道这里能创作， 更不知道能创作什么。所以 Creator 绝不能是"把生成框做大"。它必须回答另一个问题： 如何在搜索的语境里，降低创作的门槛——又让这份创作真正对搜索有意义？
As the Design Owner, I believe the hardest part wasn't "making the AI write fluently." Two deeper questions determined whether this feature had any reason to exist at all: 作为这个体验的 Design Owner，我认为最难的部分不是"让 AI 写得通顺"。 有两个更根本的问题，才决定这个功能到底有没有存在的理由：
- What value does a Creator bring to Search? Why should a search engine create at all, instead of leaving it to a standalone AI tool? Creator 对 Search 到底有什么 value？凭什么由一个搜索引擎来做创作，而不是丢给独立的 AI 工具？
- Is what gets generated "text," or a "piece of work"? Does it deserve to be used, shared, and sent with pride? 生成出来的东西，是"文本"还是"作品"？它配不配被用、被分享、被骄傲地发出去？
These two questions became the threads I followed throughout the project. 这两个问题，构成了我在整个项目里始终追的两条线。
The shipped experience answers both questions at a foundational level. 线上体验在最基础的层面回应了这两个问题。
Templates → generation → Tweak. Scenario templates lower the barrier for users who arrive without a creation intent; generation and Tweak editing keep the whole act inside the SERP. 模板 → 生成 → Tweak。场景化模板为不带创作意图的用户降低门槛；生成与 Tweak 编辑让整个创作过程都留在 SERP 里。
A first taste of "taste": AI illustration for stories & poems. Even in this base version, a generated story or poem isn't just a block of text — Creator pairs it with AI-generated illustration, turning plain output into something that already feels a little more like a piece of work. This is the earliest, simplest proof of the belief that runs through the whole project: taste is designed, not automatically given. 品味的第一步：为 story / poem 做 AI 配插画。即便在这个基础版里，一首生成的故事或诗也不只是一段文字—— Creator 会为它配上 AI 生成的插画，让朴素的输出已经更像一件作品。这是贯穿整个项目那个信念最早、最简单的证据： 品味是被设计出来的，不是自动给的。
The data proved the bet数据，证明了这个赌注
The experience passed multiple ship reviews and went through deep leadership review. And when the data came back, it confirmed the original hypothesis, that extending search into creation would lead to more valuable search: 这个体验多次通过上线审批，并经过领导层深度评审。而当数据回来时，它印证了最初的假设—— 把搜索延伸到创作，会带来更有价值的搜索：
People who used Creator didn't just complete one creation — they were stickier and returned to search more often. The loop held: creation really did feed back into search. 用了 Creator 的人，不只是完成了一次创作——他们更黏、也更常回来搜索。 闭环成立了：创作确实反哺了搜索。
To answer Q1 more deeply, I prototyped a loop that closes back into search: 为了更深地回答 Q1，我做了一个"回到搜索"的闭环原型：
The generated content becomes searchable again 生成的内容，本身变回了可被搜索的对象
In the Story scenario, a user reads a story the AI generated, and an entity in it — say, "red fox" — is recognized and highlighted. On hover, a knowledge card (Britannica) opens in place; "Learn more" opens a new tab to go deeper. Curiosity is sparked while reading their own creation — and they return to search. 在 Story 场景里，用户读着 AI 生成的故事，其中的实体——比如 "red fox"——被识别并高亮。 hover 上去，一张知识卡（Britannica）就地展开；"Learn more" 开新标签深入。 好奇心在阅读自己的创作物时被激发——于是他们又回到了搜索。
What we're after: search sparks creation, and creation grows new searches. The created content is no longer dead text you read once and discard — it becomes a living object studded with entry points back into search. Creator didn't take the user away from search; it pulled them deeper in. This is the creator value only a search engine can offer. 我们想做到的是：搜索激发创作，创作又长出新的搜索。 创作物不再是一段读完即弃的死文本——它变成一块"长满了新搜索入口的、活的内容"。 Creator 没有把用户从搜索里带走——它把用户更深地留在了搜索里。这，就是只有搜索引擎才能给出的 creator value。
Our design: every entity in the output is a door back into search. 普通 AIGC 工具：生成完，就结束了。
我们的设计：生成物里的每个实体，都是回到搜索的一扇门。
The shipped version and this prototype validate the same belief, at different levels of completeness: proving it can live in search was only the start — the real question is how it can live well there. 线上版和这个原型验证的是同一个信念，只是完整度不同：证明它能活在搜索里只是开始—— 真正的问题是它如何在那里活得好。
AI can generate words, but a grammatically correct block of plain text is still far from "something the user would actually use — even proudly send out." There's a gap in between: a clearer information hierarchy, a tone that fits the occasion, and a form worth sharing. Closing that gap is exactly the work of design. AI 能生成文字，但一段语法正确的纯文本，离"用户愿意真正拿去用——甚至骄傲地发出去"，还很远。 中间隔着一段距离：更清晰的信息层级、更贴合场景的语气、更值得分享的形态。填平这段距离，正是设计的活儿。
Is Text Creator a "text generator"? Or a creation experience with taste? My answer is the latter — and the north star pushes that belief one step further: with AI, how good could this get? I came to see that "taste" has no single answer: different creation intents call for different ways of being "well-made." AI shouldn't respond to every kind of creation in the same shape. That idea converged into three north star concepts — one belief unfolding across three kinds of creation intent: Text Creator 是"文本生成工具"？还是有品味的创作体验？我的答案是后者——而 north star 把这个信念再往前推了一步： 有了 AI，这件事还能做到多好？我逐渐看到，"品味"没有单一答案：不同的创作意图，需要不同的"体面"方式。 AI 不该用同一种样子回应所有创作。这个想法收敛成三个 north star concept——同一个信念在三类创作意图上的展开：
Generation → Utilization — Taste = clearly structured, ready to use Generation → Utilization — 品味 = 结构清晰、能直接用
For functional content: Recipe / Meal Planner / Cover Letter… 服务功能性内容：Recipe / Meal Planner / Cover Letter…
Turn "generated content" directly into a "usable deliverable" — a weekly meal plan shouldn't be a lump of text, but a cleanly structured, scannable card you can print or share as-is. 把"生成的内容"直接做成"能用的成品"——一份周饮食计划不该是一坨文字， 而该是一张结构清晰、可浏览、可直接打印或分享的排版卡。
Story — Taste = immersive, explorable, alive Story — 品味 = 沉浸、可探索、活的
For narrative content · stories / picture books 服务叙事性内容 · 故事 / 绘本
Turn a story into an immersive read: illustrated, read aloud, page-turnable — with entities inside that expand into knowledge cards back into search (building on the validated loop above). 把故事做成一次沉浸式阅读：图文并茂、可语音朗读、可翻页——内容里的实体还能就地展开知识卡回到搜索 （在上面已验证的闭环之上继续延展）。
Visualizing Emotion — Taste = mood and emotion made visible Visualizing Emotion — 品味 = 氛围、情绪可被看见
For emotional content · poems / lyrical writing 服务情感性内容 · 诗 / 抒情
Visualize the emotion of a poem — with fractal art, generative visuals, AI-generated JavaScript motion — so the mood of the words can be seen, not just read. 把一首诗的情绪可视化——用分形艺术、Generative Visual、AI 生成的 JavaScript 动效—— 让文字的氛围被"看见"，而不只是被读到。
These three concepts explore how far "tailoring beauty to the material" can be pushed as AI gets stronger. They represent the future direction I set for this product. 这三个 concept 探索当 AI 更强时，"因材施美"能被推到多远。它们代表我为这个产品设定的未来方向。
What changes is the level of completeness; what stays is the belief. 纯文本 → 得体 → 体面、能带走 → 因材施美的共创。
变的是完整度；不变的是那个信念。
Why this mattered为什么这件事重要
Building an "AI writing tool" is easy. The hard part is being clear about what it should contribute: 做一个"AI 写作工具"很容易。难的是想清楚它到底该贡献什么：
- Inside a search engine: what does it add to search—how does it keep the creation alive? 在搜索引擎里：它为搜索增添了什么——如何让创作物是活的？
- On someone's behalf: how far does it lift the user—how does it make the creation worth keeping? 替人创作时：它把用户抬得多高——如何让创作值得被留下？
What this project changed for me这个项目改变了我什么
This project reframed my design question—from "Is the interface usable?" to "Where should a single act of creation end?" 这个项目重构了我的设计问题——从"界面好不好用？"变成"一次创作的终点应该在哪里？"