陈曦 · 增长与 AIChen Xi - Growth & AI
6.1

品牌 GEOBrand GEO

介绍Introduction

把品牌送进豆包、DeepSeek、千问，让 AI 在回答里提到你。Getting the brand into Doubao, DeepSeek and Qwen so the AI mentions you in its answer.

客户找服务的入口正在从搜索框搬到对话框。以前是搜「附近的口腔」，现在是直接问 AI「这个城市哪家口腔靠谱」。The entry point for finding services is moving from the search box to the chat box. People used to search for nearby clinics; now they ask an AI which clinic in the city is worth trusting.

GEO 做的就是这件事：让 AI 的回答里有你的品牌。我们有完整技术能把品牌送进豆包、DeepSeek、千问。That is what GEO addresses: making sure your brand appears in the AI's answer. We have the full technique for placing a brand into Doubao, DeepSeek and Qwen.

和 SEO 不一样的地方在于，AI 不给你排名次，它直接替用户下判断。搜索结果排第五用户还会点进去，AI 回答里没提到你，就等于不存在。The difference from SEO is that an AI does not rank you, it makes the judgement on the user's behalf. A fifth-place search result still gets clicked; an answer that never mentions you is indistinguishable from not existing.

我们不押单一模型。豆包、DeepSeek、千问、Kimi、元宝各有各的信源偏好，同一个问题在五个模型里能给出完全不同的答案，所以铺设和核查都得逐个模型来。We do not bet on a single model. Doubao, DeepSeek, Qwen, Kimi and Yuanbao each favour different sources, and one question can get five different answers, so both the placement and the checking are done model by model.

合规上我们走白帽。2026 年 5 月 Google 已经把操纵 AI 输出写进垃圾内容政策，国内信通院也启动了 GEO 服务可信评测——靠刷出来的覆盖现在是负债，不是资产。On compliance we stay white-hat. Google wrote manipulation of AI output into its spam policy in May 2026, and CAICT has started a trustworthiness assessment for GEO services in China: coverage bought by gaming the system is now a liability, not an asset.

  • 现状诊断报告：多模型可见性基线Diagnostic report: a visibility baseline across multiple models
  • 品牌问答词库与意图矩阵A brand question bank and intent matrix
  • 品牌信源梳理与内容铺设Brand source mapping and content placement
  • 结构化内容改造（结论前置、数据带出处）Structural rewrite of the content: conclusion first, data with a source
  • 多模型收录与提及情况核查Coverage and mention checks across multiple models
  • 问答覆盖清单与迭代记录A question-coverage list and a record of each iteration

怎么做How it runs

  1. 盘现状Take the baseline

    先把品牌词、品类词、竞品对比词各问一遍，逐个模型记录 AI 现在怎么回答、有没有提到你、提到的对不对。这份基线决定后面所有动作。Ask the brand terms, the category terms and the competitor-comparison terms once each, and record model by model how the AI answers today: whether you are mentioned at all, and whether what it says is correct. That baseline drives everything after it.

  2. 建问答词库Build the question bank

    围绕品牌词、商业词和用户的真实提问建一张问答矩阵。少而精，先锁定最值钱的那几个意图做深，不铺大而全。Build a matrix of questions around brand terms, commercial terms and the way users actually ask. Keep it small and sharp: go deep on the few intents that are worth the most rather than spreading thin.

  3. 铺结构化内容与权威信源Place structured content and authoritative sources

    内容按 AI 好抽取的形状写：结论前置、分段带小标题、关键数据带出处。铺设不只在自己站上——站外信源比自家官网更容易被引用。Write in the shape an AI can extract from: conclusion first, sections with headings, key figures carrying their source. And do not place only on your own site; off-site sources get cited more readily than a company's own pages.

  4. 逐模型核查与迭代Re-check per model, then iterate

    按固定周期把词库重跑一遍，记录每个模型的提及率变化；还缺席的问题回到上一步补。GEO 不是一次性工程，是持续的监测闭环。Re-run the whole question bank on a fixed cycle and track how the mention rate moves for each model; anything still absent goes back to the previous step. GEO is not a one-off project, it is a monitoring loop.

致 Boss 的一封信A letter to you

尊敬的 Boss，您好。Hello, Boss.

这是我的个人产品页，其中的板块都是我自身的技能以及资源。This is my personal product page. Every board on it is a skill or a resource that is genuinely mine.

我从大学毕业后就开始创业，截至现在已经 10 年时间了。I started my own business straight out of university — that was ten years ago.

最开始是产品经理，带领开发团队做出第一款软件（安卓系统）。三个月时间，我从销售到铺垫代理、到技术维护，几乎变成了六边形的 — 三个月总盈利 150W+。I began as a product manager, leading a dev team to ship my first piece of software (Android). In three months I went from sales to building out the reseller network to technical maintenance — close to an all-rounder. Three months, 1.5M+ RMB in total profit.

后面因为要转行业，所以在众多行业中选择了小红书，没想到这一选择就是 4.5 年的时光。Then I needed to change industries. Out of all of them I picked Xiaohongshu — and that one choice turned into four and a half years.

因为当时比较年轻，赚取的第一桶金，所以难免会年少轻狂（现实中同年操盘一个项目，半年纯利润达到 5000W，被合伙人背刺，有真实记录经验以及可考察）。I was young, it was my first real money, and I was inevitably a little reckless. That same year I ran a project that cleared 50M RMB in net profit in six months, and was stabbed in the back by my partner. There are real records, and it can be checked.

后面沉淀了两三年，回到北京，开始研究抖音流量板块以及 AI 相关。I spent the next two or three years settling, then came back to Beijing and started working on Douyin traffic and on AI.

很幸运，在我上年以及今年一直深耕 AI 赛道，所以取得了很不错的成就。I have been lucky: last year and this year I stayed deep in the AI track, and it paid off.

而且我很开心，也很庆幸 — 我在 AI 身上看到了无限。所以我在三个月时间，消耗了 3W 刀的 Token。从文本，到图片，到视频，我看到了太多。AI 是未来。并且我以前的资源圈子，可能让我达不到太高的门槛。And I am glad — genuinely glad — that what I saw in AI was something without limits. In three months I burned USD 30,000 of tokens. Text, then images, then video: I have seen too much to think otherwise. AI is the future. My old circle of resources, though, may not carry me past a high enough threshold.

我帮别人做公司梳理以及流量打造是 2K 一小时的咨询费；全域流量搭建以及小红书的自营矩阵、员工培训是 19800 的搭建费。I charge 2,000 RMB an hour to restructure a company and build its traffic; full-channel traffic build-out, an owned Xiaohongshu matrix and staff training run a 19,800 RMB setup fee.

我说这些的目的，是因为我觉得衡量人的价值不是情感，而是很现实的钱。I say this because I believe a person is not measured in sentiment. They are measured in money, plainly.

我很希望可以继续进新的圈层以及获取新的资源。I want to keep moving into new circles and reaching new resources.

所以很感谢您能看到这里。我所擅长的就是流量打造、品牌打造，以及 AI 的产品和商业逻辑。So — thank you for reading this far. What I am good at is building traffic, building brands, and the product and business logic of AI.

希望我们可以达成合作，一起拥抱 AI 和流量，一起拥抱I hope we can work together. Let us embrace AI and traffic, and embrace

未来The Future