Knowledge Base
这里把网站上的产品信息、AI search visibility 概念、工具区别、credits、报告指标、方法论、隐私和常见排错集中到一个帮助中心里，适合第一次了解产品，也适合跑完报告后回来查具体含义。
Tools
Full Audit turns prompt-level AI observations into a deeper report with evidence, competitors, citations, and prioritized actions.
The overview gives the result a clear starting point: a time-stamped visibility score, evidence interpretation, completed sources, weighted signals, and a compact competitor snapshot.
The headline stays connected to raw counts and denominators so the user can see the scale of the evidence rather than treating one percentage as a permanent ranking.
匿名的完整报告总览。商家与竞品身份已经移除，画面保留真实的产品结构和数据关系。
查看原尺寸图片整体可见度快照
带时间戳的分数汇总本次观察中的推荐与引用证据。
基于证据的解读
报告用清楚的语言解释分数含义，以及当前最重要的可见度缺口。
已完成的 AI 平台
平台覆盖说明本次运行中哪些已选平台返回了可用证据。
可信度与对比背景
加权信号同时保留原始数量、分母和竞品出现情况，避免只看一个总分。
Each selected AI source is evaluated only on the answers it actually returned in the current run. Mention and citation counts therefore keep their own source-specific denominator.
An unavailable source is shown as not run rather than being converted into 0% performance. This keeps temporary source limitations separate from a genuine business miss.
按来源检查平台差异。截图来自真实报告，商家身份已经移除。
查看原尺寸图片按平台拆分证据
每个平台分别展示商家提及、citation signal 和竞品数量。
分母来自实际返回
比例只基于该平台在本次运行中实际返回的答案。
未运行不等于 0%
没有可用返回的平台保持 Not run，不会被误算成商家表现为零。
The prompt-by-source view shows where a citation signal was captured, where the business was missed, and where a source did not run. The three states are intentionally distinct.
In this public walkthrough, the exact prompts are replaced by high-level discovery intent labels. The real product retains the observed prompt context without exposing the private prompt set in the Knowledge Base.
Prompt × source 矩阵保留状态关系，同时隐藏具体测试措辞和品牌信息。
查看原尺寸图片公开示例保护 prompt 细节
真实 prompt 在此替换成 discovery intent 标签，不公开内部 prompt set。
Business missed 是观察结果
它表示该平台的实际答案没有出现商家，不是系统故障。
Not run 是平台状态
没有返回可用答案的平台与商家被漏掉是两种不同情况。
A citation signal does not automatically mean the business's own website was cited. Full Audit classifies cited URLs as owned, competitor, third-party, or unknown so users can see who currently supplies the evidence.
The anonymized table preserves the ownership mix, source counts, and prompt counts while replacing real domains with neutral examples.
Citation ownership 说明谁正在提供答案所依赖的证据，而不只是有没有 citation signal。
查看原尺寸图片Citation signal 不等于 owned citation
捕获 citation signal 之后，还需要判断被引用 URL 属于谁。
证据来源结构
Owned、competitor、third-party 与 unknown 分开计算，避免混为一个引用率。
URL 级证据可追溯
明细保留来源和 prompt 数量；公开示例中的真实域名已经替换。
The business and competitors use the same observed-answer denominator. Multiple entities can appear in one answer, so appearance rates are non-exclusive and may total more than 100%.
This makes the comparison useful without presenting it as traditional share of voice. The purpose is to show who repeatedly occupies the answer space when the business appears or is absent.
匿名的竞争视图保留真实出现次数和分母，同时移除品牌与竞品身份。
查看原尺寸图片统一分母
商家和所有竞品都使用相同的 observed-answer 分母，方便公平比较。
出现率可以重叠
一个答案可以出现多个实体，因此总和可能超过 100%。
区分被搜索商家和观察到的竞品
表格说明实体角色与出现次数，不把它包装成传统 SOV 排名。
The Action Plan keeps one shared priority order and connects each recommendation to the affected page, the observed intent, the relevant sources, and the evidence seen in the audit.
The public example generalizes the business-specific wording and URLs, while preserving impact, effort, target location, and the ability to inspect supporting detail.
公开示例将业务措辞与 URL 泛化，但保留真实的优先级和证据追踪结构。
查看原尺寸图片一套共用优先级
所有建议按照 Do first 与 Next 排序，避免不同页面出现互相冲突的顺序。
保留 impact 与 effort
每条建议在标题层就能看出预期影响和执行成本。
明确目标页面
建议连接到需要修改或创建的页面位置。
证据可以继续展开
用户可以查看相关 intent、来源和 audit evidence，而不是只读一句建议。
Full Audit shows where the brand is visible, where it is missing, who replaces it, which kinds of pages are cited, and what to fix first. The evidence is tied directly to observed AI answers rather than a generic SEO checklist.
The result is a traceable path from overview to evidence, comparison, and action.