[{"data":1,"prerenderedAt":397},["ShallowReactive",2],{"blog-china-site-trust-signals":3,"blog-prev-china-site-trust-signals":90,"blog-next-china-site-trust-signals":289},{"id":4,"title":5,"author":6,"body":7,"canonical":74,"category":75,"cover":74,"date":76,"description":77,"draft":78,"extension":79,"keywords":74,"meta":80,"navigation":81,"noindex":78,"ogImage":74,"path":82,"seo":83,"seoDescription":74,"seoTitle":74,"stem":84,"tags":85,"updated":74,"__hash__":89},"blog\u002Fblog\u002Fchina-site-trust-signals.md","中国站点信任信号清单：备案、工商信息和真人","元析智能",{"type":8,"value":9,"toc":67},"minimark",[10,14,18,25,31,37,43,46,49,52,55,58,61,64],[11,12,13],"p",{},"国内用户判断一个软件站可不可信，看的是几个具体信号，跟设计好不好看关系不大。2025 年到现在我们给 40 多个国内软件站做过信任页评审，四项基础信息不全的占了六成。",[15,16,17],"h2",{"id":17},"四项必备",[11,19,20,24],{},[21,22,23],"strong",{},"ICP 备案号。"," 放在页脚，链接到工信部备案查询页。查法很简单，去工信部系统输入域名，能查到备案主体和备案号。页面上的号必须和查询结果一致，过期、注销、写成别家公司的，都算不合格。",[11,26,27,30],{},[21,28,29],{},"公安备案。"," 通常和 ICP 并列放在页脚，带公安图标和编号。很多站只做了 ICP 没做公安，严格来说是不完整的。",[11,32,33,36],{},[21,34,35],{},"工商信息。"," 公司全称、统一社会信用代码、注册地址。至少要在关于页能查到。用户真要核实，会去国家企业信用信息公示系统搜。",[11,38,39,42],{},[21,40,41],{},"公开联系方式。"," 一个能打通的电话，一个能到门牌号的地址。这两项看着简单，实际最常出问题：电话是空号、地址只写到城市、留的是虚拟办公地址。",[15,44,45],{"id":45},"三个加分项",[11,47,48],{},"真实的办公照片或短视频，比任何文字都管用。团队页放真人姓名和职位，别用卡通插画顶替。客户 logo 墙要拿到授权，没授权就写行业和规模，别放 logo。",[15,50,51],{"id":51},"常见错误",[11,53,54],{},"最常见的是备案主体和运营主体不一致。2025 年 8 月我们评审一个客户，备案主体是个人，运营方是公司，企业客户在采购流程里直接提出质疑，最后花了两周重新备案。",[11,56,57],{},"第二常见的是电话打不通。我们随机拨了 20 个客户的公开电话，有 3 个是空号，2 个转到了语音信箱且没回电。",[11,59,60],{},"第三是地址变更没同步。公司搬了家，页脚还是老地址，客户按地址找过去是别家公司，这种信任损失很难补。",[11,62,63],{},"还有一个隐性问题是备案号被放在图片里。有些站把备案号做成一张小图，看着整齐，但用户和工具都读不到，也点不进工信部查询页。改成文字加链接，两分钟的事。",[11,65,66],{},"这份清单两个小时内能查完，但六成站点不及格。都是些小事，只是没人认真做。我们现在的做法是把这张清单放进新站上线的检查表，逐项有人签字，免得拖到被客户问起才补。",{"title":68,"searchDepth":69,"depth":69,"links":70},"",2,[71,72,73],{"id":17,"depth":69,"text":17},{"id":45,"depth":69,"text":45},{"id":51,"depth":69,"text":51},null,"技术SEO","2026-03-31","我们给 40 多个国内软件站做过信任页评审，ICP 备案、公安备案、工商信息、公开电话四项不全的占六成。这篇是一份可逐项打勾的清单，每项都写了怎么查、查到什么算过关。",false,"md",{},true,"\u002Fblog\u002Fchina-site-trust-signals",{"title":5,"description":77},"blog\u002Fchina-site-trust-signals",[86,87,88],"备案","工商信息","站点可信度","4cV3lE2YBlcvujmiNILjk_XLfAb2vWv0iBPyy6JjASw",{"id":91,"title":92,"author":6,"body":93,"canonical":74,"category":277,"cover":74,"date":278,"description":279,"draft":78,"extension":79,"keywords":74,"meta":280,"navigation":81,"noindex":78,"ogImage":74,"path":281,"seo":282,"seoDescription":74,"seoTitle":74,"stem":283,"tags":284,"updated":74,"__hash__":288},"blog\u002Fblog\u002Fbrand-corpus-audit.md","品牌语料盘点：五个客户平均 240 个来源，官网只占 30 个",{"type":8,"value":94,"toc":268},[95,98,101,104,107,110,113,116,119,122,211,214,217,220,223,226,229,232,235,238,241,244,247,250,253,256,259,262,265],[11,96,97],{},"AI 认识你，靠的是 240 个页面，其中只有 30 个是你写的。2026 年 3 月我们给五个 B2B 软件客户做了语料盘点，把四个引擎在品牌词和产品词问答里挂出来的来源全部列出来、去重、分类。平均 240 个来源，自家官网占 30 个，12.5%。",[15,99,100],{"id":100},"盘点要回答三个问题",[11,102,103],{},"它读了什么：哪些页面在给 AI 提供关于你的信息。\n它读错了什么：哪些页面的描述和事实不符。\n它缺了什么：哪些信息本该有、但一个来源都没有。",[11,105,106],{},"这三个问题比\"我们被引用了几次\"有用得多。次数是结果，语料是原因。",[15,108,109],{"id":109},"怎么收集来源",[11,111,112],{},"四个引擎，每个跑 20 个查询：10 个品牌词，5 个产品词，5 个\"品牌加场景\"。",[11,114,115],{},"每个回答记录两样东西：挂出来的域名，以及它描述你的原话。原话必须抄，不能只记域名。",[11,117,118],{},"五个客户、400 次问答，去重后拿到 1180 个来源域名，平均每个客户 236 个。顺手还收到 300 多条描述原话，这批原话是后面判断对错的依据。",[15,120,121],{"id":121},"平均结构长什么样",[123,124,125,141],"table",{},[126,127,128],"thead",{},[129,130,131,135,138],"tr",{},[132,133,134],"th",{},"来源类型",[132,136,137],{},"平均数量",[132,139,140],{},"占比",[142,143,144,156,167,178,189,200],"tbody",{},[129,145,146,150,153],{},[147,148,149],"td",{},"官网",[147,151,152],{},"30",[147,154,155],{},"13%",[129,157,158,161,164],{},[147,159,160],{},"平台内容（知乎、公众号、掘金、CSDN）",[147,162,163],{},"96",[147,165,166],{},"40%",[129,168,169,172,175],{},[147,170,171],{},"媒体与自媒体",[147,173,174],{},"42",[147,176,177],{},"18%",[129,179,180,183,186],{},[147,181,182],{},"目录站与工具导航",[147,184,185],{},"38",[147,187,188],{},"16%",[129,190,191,194,197],{},[147,192,193],{},"GitHub 与技术社区",[147,195,196],{},"22",[147,198,199],{},"9%",[129,201,202,205,208],{},[147,203,204],{},"其他",[147,206,207],{},"10",[147,209,210],{},"4%",[11,212,213],{},"平台内容占了四成，这个数字每次都让客户惊讶。他们一直以为官网是主战场，实际主战场在别人的地盘上。",[15,215,216],{"id":216},"读错了什么",[11,218,219],{},"五个客户里有三个存在明显错误描述。",[11,221,222],{},"一个被写成\"仅支持私有部署\"，实际有 SaaS 版。错误来源是 2023 年的一篇媒体报道，被三个目录站转载。",[11,224,225],{},"一个被写成\"开源软件\"，实际只有 SDK 开源。来源是 GitHub 上一个第三方项目的 README。",[11,227,228],{},"一个被写成\"总公司在深圳\"，实际在成都。来源是某招聘网站的旧信息。",[11,230,231],{},"三条错误的共同点：都不是官网写的，但都被 AI 当成事实，因为多个来源重复了同一个说法。",[15,233,234],{"id":234},"缺了什么",[11,236,237],{},"缺口集中在三块：中文场景内容、新版本文档、客户案例的量化结果。",[11,239,240],{},"新版本文档缺得最要命。一个客户 2025 年上线了三个新功能，文档站更新了，但中文语料里一个字都没有。结果 AI 回答\"你们支持向量检索吗\"时，说\"暂不支持\"。",[11,242,243],{},"这条错误每天在替他拒绝客户。",[15,245,246],{"id":246},"盘点之后做什么",[11,248,249],{},"三类动作，按优先级：",[11,251,252],{},"改错的。找到错误来源页，发更正请求。改不动的，就在官网和平台写一版更清楚的说法，让正确信息在数量上压过错误信息。",[11,254,255],{},"补缺的。缺口里最影响成单的先补，通常是新版本文档和部署方式。",[11,257,258],{},"建台账的。把 240 个来源记成一张表，标注类型、是否准确、最后检查日期。每半年复查一次。",[15,260,261],{"id":261},"一个判断",[11,263,264],{},"语料盘点比关键词研究有用。关键词研究告诉你别人搜什么，语料盘点告诉你 AI 以为你是谁。后者错了，前者做得再好也没用。",[11,266,267],{},"这活儿一次四个小时，五个人工。不做，你就一直在猜。",{"title":68,"searchDepth":69,"depth":69,"links":269},[270,271,272,273,274,275,276],{"id":100,"depth":69,"text":100},{"id":109,"depth":69,"text":109},{"id":121,"depth":69,"text":121},{"id":216,"depth":69,"text":216},{"id":234,"depth":69,"text":234},{"id":246,"depth":69,"text":246},{"id":261,"depth":69,"text":261},"GEO","2026-03-30","我们把 AI 能读到的关于品牌的页面全列出来，五个客户平均 240 个来源，自家官网只有 30 个。盘点要回答三个问题：它读了什么、读错了什么、缺了什么。",{},"\u002Fblog\u002Fbrand-corpus-audit",{"title":92,"description":279},"blog\u002Fbrand-corpus-audit",[285,286,277,287],"语料","品牌实体","盘点","-XSdAAZ8qz0gjHInPI5tuRSI7rtw_Kmek-os-70Xwgs",{"id":290,"title":291,"author":6,"body":292,"canonical":74,"category":75,"cover":74,"date":386,"description":387,"draft":78,"extension":79,"keywords":74,"meta":388,"navigation":81,"noindex":78,"ogImage":74,"path":389,"seo":390,"seoDescription":74,"seoTitle":74,"stem":391,"tags":392,"updated":74,"__hash__":396},"blog\u002Fblog\u002Fcross-border-software-hreflang.md","跨境软件做多语言站，hreflang 错了等于白做",{"type":8,"value":293,"toc":380},[294,297,300,303,306,331,334,337,340,343,365,368,371,374,377],[11,295,296],{},"跨境软件的多语言站，十个里有六个 hreflang 配错。2026 年 2 月我们一次性看了三个客户的配置，两个是错的，一个是半错的。错法的后果不是惩罚，是各语言版本互相抢排名，或者该出现的版本不出现。这类问题在 Search Console 里看不出来，得手动验。",[15,298,299],{"id":299},"三种常见错法",[11,301,302],{},"第一种，单向标注。英文页指向了德文页，德文页没指回英文页。hreflang 必须双向，你指我我指你，单向等于没配。",[11,304,305],{},"第二种，x-default 乱放。x-default 是给“没有匹配语言”的用户看的版本，一般放英文或者语言选择页。有人把它放在主站首页，导致所有语言版本都在争同一个默认。",[11,307,308,309,313,314,317,318,321,322,324,325,327,328,330],{},"第三种，语言代码和地区代码混用。",[310,311,312],"code",{},"de"," 是德语，",[310,315,316],{},"de-DE"," 是德国德语，",[310,319,320],{},"de-AT"," 是奥地利德语。跨境业务面向德国和奥地利，用 ",[310,323,312],{}," 覆盖不了地区差异；但如果你只有一个德语版本，写 ",[310,326,312],{}," 就够，写 ",[310,329,316],{}," 反而把奥地利用户排除在外。",[15,332,333],{"id":333},"怎么验证",[11,335,336],{},"三个动作。一是拿 curl 抓每种语言版本，看返回的 HTML head 里有没有完整的 alternate 标签组。二是核对每个页面的 hreflang 组是否完整，五语言站应该有五条互指加一条 x-default。三是用 Search Console 的国际化报告看有没有报错，注意这个报告只报冲突，不报缺失。",[11,338,339],{},"我们做这套检查写了个脚本，跑一遍两分钟，比人工点省事。需要的话可以照着我们说的逻辑自己写，核心就是解析 head 里的 link 标签，检查组内互指。",[15,341,342],{"id":342},"域名结构怎么选",[11,344,345,346,349,350,353,354,357,358,360,361,364],{},"三种选法：语言放子目录（",[310,347,348],{},"\u002Fde\u002F","）、放子域名（",[310,351,352],{},"de.example.com","）、用独立域名（",[310,355,356],{},"example.de","），都是一个语言版本一套结构，别混着来。我们一般推荐 ",[310,359,356],{}," 这种。它的优势是本地用户信任度高，德国用户看到 ",[310,362,363],{},".de"," 会觉得你是本地公司；代价是每个域名都要单独积累权重，起量慢。",[11,366,367],{},"预算有限时选子目录，权重集中在主域，见效快。子域名是最不推荐的，权重独立又共享不了品牌，多语言站做子域名的，十个有八个在半年后重新合并。",[15,369,370],{"id":370},"什么情况别分语言站",[11,372,373],{},"产品只有英文界面，硬做德语页，用户点进去发现界面还是英文，跳出率高，还拉低整站质量。这种情况下正确的做法是一个英文站，用内容页承接搜索，别做假本地化。",[11,375,376],{},"另一个情况是语言版本内容量差太多。德文站只有 5 个页面，英文站 300 个，搜索引擎会认为德文站不完整，两边的排名都受影响。要么补齐，要么先别开。",[11,378,379],{},"跨境软件的多语言 SEO，工程细节比内容重要。hreflang 只是其中一环，还有货币、域名结构、CDN 区域。这些小地方错一个，流量就少一截，而且很难从报表里发现问题。",{"title":68,"searchDepth":69,"depth":69,"links":381},[382,383,384,385],{"id":299,"depth":69,"text":299},{"id":333,"depth":69,"text":333},{"id":342,"depth":69,"text":342},{"id":370,"depth":69,"text":370},"2026-04-01","我们接手过三个跨境 SaaS 的多语言站，两个的 hreflang 配置是错的。常见的三种错法、怎么用 Search Console 验证，以及什么情况干脆别分语言站。",{},"\u002Fblog\u002Fcross-border-software-hreflang",{"title":291,"description":387},"blog\u002Fcross-border-software-hreflang",[393,394,395],"hreflang","多语言站","跨境软件","Pe7Hkwo52wkqERrpjrg04-X1jmNmqT27gfmUNKN-8j4",1794088107161]