SuperAPI AI Directory

geo-tracking-plan

General & other

View repository

Updated 10-16 09:30

About this skill

GEO Tracking Plan

When To Use

  • Input the company name, brand name, and a small amount of supporting information, and expect the system to complete authoritative retrieval on its own and generate a GEO backend effect attribution framework and tracking system plan suitable for this company.
  • It is necessary to use the official website as the core basis, combine public materials and user-supplied information, identify business characteristics, and then output a personalized monitoring design for direct effects and indirect effects.
  • The deliverable needs to be a structured plan, and ideally it should also directly generate HTML and Word files.
  • It is necessary to map "enterprise type, conversion path, existing site capabilities, trackable signals" into an implementation roadmap with clear priorities.

Do Not Use

  • The user only wants to understand what GEO is, or only wants a set of general methodology popularization.
  • The user only wants to do GEO content strategy, topic planning, source analysis, or competitor research.
  • The user only wants to do SEO, SEM, CRM tracking, or BI dashboard implementation, and does not need GEO attribution framework design.
  • The user explicitly wants only code implementation, not analysis plans, data definitions, or attribution design.

Required Inputs

The minimum input is only the company name. If the user provides the following information, absorb it as well:

  • Brand aliases, English name, product names, core business lines
  • Known official website or suspected official website links
  • Target market, customer groups, typical conversion actions
  • Existing landing pages, forms, hotlines, WeCom, promo codes, consultant entry points, and other reception assets
  • Key AI platforms, key keyword topics, core actions to be monitored
  • Whether direct output is needed HTML / Word

If the user has not provided supplementary information, still start from the company name and explicitly list the information gaps at the end.

Required Reading

Read as needed before starting:

  • references/evidence-sourcing-workflow.md
  • references/attribution-framework.md
  • references/personalization-matrix.md
  • references/market-split-framework.md
  • references/quality-gates.md
  • references/html-output-outline.md
  • references/report-ui-rules.md
  • references/output-file-workflow.md
  • references/open-source-sanitization.md
  • references/observability-estimation-framework.md

Workflow

  1. First unify the enterprise entity.
    • Standardize the company's full name, brand name, product names, aliases, and English name.
    • If there is more than one company with the same name, first perform candidate alignment, then continue the analysis.
  2. Do not use internal systems as a default dependency.
    • Prioritize information provided by the user and publicly verifiable materials.
    • If internet access is available, first confirm the official domain, then supplement with public evidence such as the official website, product pages, about page, contact page, pricing page, case studies page, and careers page.
    • If internet access is unavailable, explicitly state that only a low-confidence version based on user-provided materials is being output.
  3. Determine the market scope and platform environment.
    • Clarify whether the current analysis is more oriented toward 国内 GEO、海外 GEO or 混合 GEO。
    • If domestic GEO mainly revolves around DeepSeek、豆包、元宝、Kimi and other platforms, increase the weight of passphrases, questionnaires, WeCom, phone calls, campaign pages, and dedicated reception actions.
    • If overseas GEO mainly revolves around ChatGPT、Gemini、Perplexity and other platforms, you can rely more strongly on official website content structure, landing pages, form fields, web analytics, and the official page matrix.
  4. Establish a public evidence ledger.
    • Record every key piece of material source_system、source_locator、absolute_time、fact_or_inference、how_used。
    • Conclusions related to "latest" must include an absolute date; if the date cannot be obtained, do not say "latest."
  5. First determine the business type and conversion path.
    • At minimum determine B2B 销售型、PLG / SaaS、电商 / 标准化消费、预约 / 咨询型服务 the primary type among them.
    • Combine references/personalization-matrix.md to select the most matching reception actions and metric definitions.
  6. Verify the official website, and do not treat secondhand pages as the official website.
    • Verify the consistency of the brand's self-description, About Us, contact information, legal pages, social media, or careers pages.
    • Do not mistake media releases, directory sites, or agency pages for the official website.
  7. Establish a backend analysis draft for enterprise GEO.
    • At minimum, record the company positioning, core products, target audience, typical conversion paths, existing承接 assets, current monitoring capabilities, and site capability limitations.
    • Clarify which are verified facts and which are inferences based on facts.
  8. Model in two layers: "direct effects / indirect effects."
    • First break down GEO value into 品牌层价值 and 效果转化层价值。
    • Brand-layer value emphasizes the role of "third-party endorsement" on AI platforms, including brand awareness, trust building, mindshare reinforcement, and reduced friction in subsequent conversions.
    • The effect conversion layer value is further broken down into 可直接监测的效果 and 间接促进的效果。
    • Direct effects: dedicated phone number, dedicated mobile number, dedicated WeChat ID, dedicated advisor entry point, dedicated benefit code, AI-exclusive landing page, form fields, source parameters, campaign page, customized CTA.
    • Indirect effects: branded search volume, official website UV, landing page UV, CTA click-through rate, registration rate, appointment rate, lead rate, conversion rate, self-reported source questionnaire.
    • Under 现状诊断 , a 效果追踪方法与原理说明 module must be added, using a table to list the principles, methods, applicable conditions, and recommended execution actions for all direct and indirect monitoring; if the deliverable is HTML, preferably also include a set of process diagrams.
    • Under 间接效果追踪方案 , a 监测效果边界说明 module must be added, explaining that directly observable GEO effects are usually only part of the total contribution, and a common conservative planning assumption is about 20%~30%, and may be higher in scenarios with strong overseas official website承接; this ratio can only serve as a planning and interpretation boundary and must not be written as a fixed cross-industry fact.
    • Under 监测效果边界说明 , a 可观测性估算框架 module must continue to be added, outputting Observed / Recoverable / Unobservable three-layer planning values or ranges, applicable premises, main determining factors, and next-stage calibration actions.
    • Clearly state: the total GEO effect usually cannot be fully tracked, and the focus of the plan is to establish a minimal closed loop sufficient to explain trends, calibrate direction, and guide optimization actions.
  9. Select a plan based on the company's actual situation and market type.
    • For high-ticket, long-decision-chain businesses, prioritize lead attribution, dedicated advisor entry points, form fields, and supplementary questionnaires.
    • For standardized consumer businesses, prioritize landing pages, discount codes, source parameters, registration conversion, and post-purchase self-reported source.
    • Overseas GEO can be designed more around official website page matrices, form sources, page events, and upgrade paths.
    • Domestic GEO usually needs to place more emphasis on codes, campaign pages, WeCom/phone, questionnaire supplementation, and cross-device source patches.
    • If the official website capability is weak, prioritize supplementing承接 pages and a minimal attribution closed loop, rather than piling on complex metrics.
  10. Output a personalized plan.
  • At minimum cover objectives, definitions, data sources, implementation methods, priorities, responsible parties, explainability, gaps, and risks.
  • If the user wants an "analysis plan" rather than an "implementation plan," also provide a minimal actionable checklist.
  1. If the user requests a deliverable file.
  • First organize the conclusions into report_input.json。
  • then run:
  • python3 scripts/render_geo_tracking_plan.py --input <report_input.json> --output-dir <输出目录>
  • By default, generate .html and .docx。
  1. The public version skill must remain privacy-clean.
  • Do not require any private CLI, intranet address, private table, or internal company document.
  • Do not put private company samples, undisclosed metrics, or internal screenshots into examples, assets, reports, or delivery examples.

Output Contract

Each execution must return at least the following:

  • 企业理解摘要: company positioning, products, target audience, typical conversion actions
  • 市场分层判断: domestic / overseas / hybrid GEO and its impact on the plan
  • 公开证据表: sources, absolute dates, key facts, and uses of the official website and public materials
  • 官网核验: official domain, verification basis, and related public pages
  • 现状诊断: existing GEO reception and attribution capabilities, obvious gaps, and available assets
  • 效果追踪方法与原理说明: brand value, conversion value, direct and indirect monitoring logic, including at least one methodology table and one set of diagrams or process structures
  • 直接效果追踪方案
  • 间接效果追踪方案
  • 监测效果边界说明
  • 可观测性估算框架
  • 归因口径与数据表设计
  • 优先级路线图
  • 置信度与缺口
  • 可视化 HTML 方案
  • 可交付文件: if requested by the user, actually generate .html and .docx

Validation Checklist

  • The output does not rely on any private system, private API, internal knowledge base, or unauthorized materials.
  • The official website has been explicitly verified rather than subjectively determined based on search results.
  • The domestic / overseas / hybrid GEO scenario has been judged, and the focus of the monitoring method has been adjusted accordingly.
  • Each key judgment can be traced back to the evidence table, and facts are distinguished from inferences.
  • All claims related to "latest" include absolute dates.
  • The output clearly distinguishes direct effects and indirect effects.
  • The output clearly distinguishes brand-level value, direct effects, and indirect effects, and explains the relationship among the three.
  • The output includes 效果追踪方法与原理说明、监测效果边界说明 as well as 可观测性估算框架。
  • If outputted Observed / Recoverable / Unobservable for the three-layer percentage allocation, the three must add up to 100%。
  • The plan reflects differences in enterprise business type, conversion path, and site capabilities, rather than generic template boilerplate.
  • 20%~30% If observable proportions such as are used, they must be written as conservative planning assumptions or boundary explanations, and must not be passed off as universal conclusions.
  • If public evidence is insufficient, explicitly state the information gaps and the reasons for reduced certainty.
  • If the user requests deliverables, actually generate .html and .docx files, rather than only saying they will be generated.

Reference Map

  • references/evidence-sourcing-workflow.md: public evidence acquisition order and evidence table specifications
  • references/attribution-framework.md: direct effects, indirect effects, and data design framework
  • references/personalization-matrix.md: personalized selection of attribution actions by business type
  • references/market-split-framework.md: monitoring differences and routing rules between domestic GEO and overseas GEO
  • references/quality-gates.md: quality gates for the open-source version skill
  • references/html-output-outline.md: HTML report structure and visualization modules
  • references/report-ui-rules.md: UI rules for designing a systematic HTML report
  • references/output-file-workflow.md: report input structure and file generation process
  • references/open-source-sanitization.md: rules for removing privacy, removing intranet dependencies, and replacing with public examples
  • references/observability-estimation-framework.md: three-layer observability estimation framework and interval rules
  • scripts/render_geo_tracking_plan.py: generate HTML / Word deliverables
  • examples/hubspot-demo/report_input.json: public demo input examples
  • evals/trigger_cases.json: trigger boundaries
  • evals/quality_cases.json: quality and privacy gate examples