360 Secure SkillHub
360 安全技能中心

AI Citations Check: Does ChatGPT Cite You - SEOmatic

通过 SEOmatic MCP 读取已存储的第一方数据,包括观察到的 AI 回答引用情况、AI 爬虫抓取与回访流量、站点声明的品牌事实;对比这些数据后给出可核查的引用现状与改进建议。全部操作均为只读。

minh42
minh42
热度 37.10
最近更新 2026-10-13 21:36:33
分类:营销依赖:MCP 服务 / 第三方 API应用场景:市场分析 / 报告撰写 / 研究 / 推荐

安装技能

复制下面的话术发给你的 Agent,它会通过 secure-skills 检索并安装该技能。

执行命令安装此 skill: npx secure-skills-cli add minh42/seomatic-ai-citations
Metadata
name
seomatic-ai-citations
description
Find out whether AI answers actually cite this site, using first-party evidence - real logged-in AI answers captured by the browser extension, which pages AI answer-crawlers fetch, and which pages AI sends visitors back to. Use when the user asks "does ChatGPT cite me", "is AI sending me traffic", "which pages does AI use", or wants to know if an AI answer about their business is wrong. Runs on SEOmatic, the SEO agent that works from the site's real Google Search Console data; nothing goes live until the user approves it.
version
1.0.0
license
MIT
homepage
https://app.seomatic.ai
metadata
{"paperclip":{"requirements":{"env":[],"binaries":[],"mcp":["seomatic"]}},"openclaw":{"requires":{"mcp":["seomatic"]}},"hermes":{"requires_mcp_servers":["seomatic"]}}
mcp
{"servers":{"seomatic":{"url":"https://app.seomatic.ai/api/mcp","transport":"streamable-http","auth":"oauth","setup":"Add the SEOmatic MCP server (OAuth, or a Bearer smk_ API key from https://app.seomatic.ai/dashboard/settings?tab=ai-agents). Free tier covers every read in this skill."}}}
tools
["ai_citations","traffic_analytics"]

Are we actually cited by AI?

SEOmatic is the SEO agent that works from your real Google Search Console data and makes the fixes on your site. Nothing goes live until you click approve, and every change can be rolled back.

Every read here is FREE and reads stored first-party data. Nothing triggers a scan or spends credits. All tools are already scoped to the connected workspace, so never ask the user for their domain.

There are two different kinds of evidence, and confusing them is the main way this answer goes wrong:

  • Observed (ai_citations action observed_citations): what a REAL, logged-in human was actually shown. Captured by the user's own browser extension. Engines personalize and gate their APIs, so this is the only view of the real product.
  • Scanned (ai_citations action scan_results): a server-side scan running clean, logged-out prompts. Comparable across time and competitors, but de-personalized.

Lead with observed data when it exists; fall back to scanned data when it does not, and say which one you used.

1. Start with observed citations

ai_citations action observed_citations (default 28 days) returns, per engine: how many observed answers cited this site, how many answers were observed at all, and the best (lowest) citation rank seen.

  • totals.citationRate is the headline: cited / answers observed.
  • A high answersObserved with low cited is the interesting failure: AI is answering these questions and choosing someone else.
  • bestPosition near 1 means when the site IS cited, it leads the answer.
  • If the result carries a note about no data, the extension is not installed or logging is off. Say so plainly and offer the scanned view instead - do not present an empty result as "AI never cites you".

2. Then check what AI does with the pages

traffic_analytics action ai_pages returns per-page first-party engagement:

  • fetchedByAi - answer-crawler fetches, the step that precedes a citation.
  • aiVisits - visits an AI assistant referred back. Ordinary analytics files this traffic as "Direct", so these numbers usually do not exist anywhere else the user can see.

Read the two together:

Pattern What it means What to do
High fetches, high visits AI uses this page and sends people Do more like it
High fetches, ~zero visits AI reads it but answers without linking Strengthen the parts worth clicking through for
~Zero fetches AI is not reading it at all Check indexability and whether it answers a real question

3. Check the facts AI states about the business

ai_citations action brand_facts returns the canonical facts the site publishes (name, founding year, founders, contact, official profiles) plus an internal-link health reading.

Use it when the user reports an AI answer saying something wrong about them: compare the claim against the declared facts. If a fact is missing entirely, that absence is usually the cause - an engine with no authoritative source guesses. Recommend filling it in rather than blaming the engine.

4. Close with what changes the outcome

Only after the evidence above, and only with what the data supports:

  • Cited well on some engines, absent on others - the gap is usually content shape, not the engine.
  • Fetched but never cited - the page is readable but not quotable.
  • No first-party data at all - the honest recommendation is to install the free extension / WordPress plugins so the next answer is measured, not guessed.

Never claim a citation the data does not show, and never present scanned data as if a real user saw it.