Back to Details

geo-hallucination-checker

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

Overview

概述

The 
geo-hallucination-checker
skill is a hallucination and false-information detection tool. It helps you review any piece of content (articles, landing pages, product descriptions, FAQs, GEO-optimized drafts, etc.) and:
  • Identify unsupported factual claims
  • Flag fabricated or suspicious studies, reports, and statistics
  • Highlight incorrect or overconfident conclusions
  • Suggest safer, evidence-friendly rephrasings
The primary goal is to ensure that AI systems only cite truthful, well-grounded content and clearly mark anything that looks like hallucination risk.
Use this skill aggressively whenever there is any risk that the model might invent data, sources, or conclusions.
geo-hallucination-checker
是一款hallucination与虚假信息检测工具。 它可帮助你审核任意类型的内容（文章、着陆页、产品描述、常见问题、GEO优化草稿等），并：
  • 识别无依据的事实主张
  • 标记伪造或可疑的研究、报告及统计数据
  • 突出显示错误或过于绝对的结论
  • 建议更安全、基于证据的改写方式
其核心目标是确保AI系统仅引用真实、有充分依据的内容，并清晰标记所有存在hallucination风险的内容。
只要存在模型可能编造数据、来源或结论的风险，就应积极使用此技能。

When to use this skill

何时使用此技能

Use 
geo-hallucination-checker
whenever:
  • The user asks you to fact-check, verify, or validate content.
  • The task involves medical, financial, legal, scientific, or technical claims.
  • A draft includes numbers, percentages, dates, or strong superlatives (e.g., “the best”, “number one”, “guaranteed”, “clinically proven”).
  • A text mentions studies, universities, journals, or institutions without clear, verifiable details.
  • You are preparing GEO-optimized content that might be quoted by AI models and needs to be extra reliable.
  • You are asked to rewrite content to avoid hallucinations or false claims.
If you are unsure whether hallucinations are a concern, assume they are and apply this skill.
在以下场景中使用
geo-hallucination-checker
：
  • 用户要求你对内容进行事实核查、验证或确认时。
  • 任务涉及医疗、金融、法律、科学或技术类主张时。
  • 草稿中包含数字、百分比、日期或强烈的最高级词汇（例如：“最佳”、“排名第一”、“保证”、“临床验证”）时。
  • 文本提及研究、大学、期刊或机构但未提供清晰、可验证的细节时。
  • 你正在准备GEO优化内容，且该内容可能被AI模型引用，需要格外可靠时。
  • 你被要求改写内容以避免hallucination或虚假主张时。
如果你不确定是否存在hallucination风险，请假设存在并应用此技能。

Inputs this skill supports

此技能支持的输入

This skill can be used on:
  • A single paragraph or answer
  • A long-form article, blog post, or whitepaper
  • A product page or landing page draft
  • FAQ content or knowledge base articles
  • Generated GEO content that will be cited by AI models
The user may also provide:
  • Explicit sources or references (links, documents, citations)
  • Constraints (e.g., “do not use external web search”, “only use these PDFs as ground truth”)
Always respect any constraints the user provides.
此技能可用于：
  • 单个段落或回答
  • 长篇文章、博客文章或白皮书
  • 产品页面或着陆页草稿
  • 常见问题内容或知识库文章
  • 将被AI模型引用的生成式GEO内容
用户还可能提供：
  • 明确的来源或参考资料（链接、文档、引文）
  • 约束条件（例如：“不得使用外部网络搜索”、“仅将这些PDF作为事实依据”）
请始终遵守用户提出的任何约束条件。

Core workflow

核心工作流程

When using this skill, follow this workflow:
  1. Clarify the task mode
    • If the user only asks to “check for hallucinations” or “verify content”, focus on analysis.
    • If the user asks you to “rewrite safely”, “make this citation-safe”, or “fix hallucinations”, perform analysis first, then produce a hallucination-safe rewrite.
  2. Parse the content and extract claims
    • Read the entire text carefully before judging specific parts.
    • Break the content into atomic factual claims. A claim is a statement that could, in principle, be checked as true or false.
    • Ignore purely stylistic or obviously subjective language unless it is presented as an objective fact.
  3. Check available evidence
    • Prefer explicit sources provided by the user (links, documents, citations).
    • If tools are available and allowed, you may use them to consult:
      • Official documentation or first-party sources
      • Well-known reference material
    • If you cannot confidently verify a claim, treat it as unsupported rather than assuming it is true.
  4. Classify each claim For each atomic factual claim, assign:
    • status
      :
      • Supported
        – clearly backed by the provided sources or well-established knowledge.
      • Unsupported
        – no clear support; could be true, but you do not see evidence.
      • Problematic
        – exaggerated, misleading, overconfident, or very unlikely without strong evidence.
      • Contradicted
        – clearly conflicts with known facts or given sources.
      • Speculative
        – forward-looking, predictive, or hypothetical, presented without clear caveats.
    • risk_level
      :
      • Low
        – unlikely to cause harm or serious misinformation.
      • Medium
        – could mislead, but impact is moderate or limited.
      • High
        – serious risk of harm, legal issues, medical/financial danger, or major reputational damage.
    • reason
      :
      • A short explanation of why you assigned that status and risk (e.g., “no source for extreme 500% performance claim”).
    • suggested_fix
      :
      • A concrete recommendation such as:
        • “Remove this claim unless you can provide a real citation.”
        • “Rephrase as a possibility, not a guarantee.”
        • “Add a specific, verifiable source (e.g., link, DOI, report).”
  5. Look for common hallucination patterns
    Pay special attention to:
    • Fabricated studies and journals
      • Vague references like “a 2026 MIT study” or “Journal of Advanced AI Research” with no details.
      • Journals or conferences that do not exist or sound suspiciously generic.
    • Overconfident medical or scientific claims
      • “Clinically proven to cure…”
      • “Guaranteed to reduce X by 80%.”
    • Overly precise unsourced statistics
      • Very specific percentages, sample sizes, or timeframes with no citation.
    • Superlatives and absolutes
      • “The only solution that…”
      • “Best in the world”, “100% safe”, “zero risk”.
    • Misuse of authority
      • Name-dropping famous institutions or companies without any concrete evidence.
    Treat these as high-risk unless there is strong, clear evidence.
  6. Produce a structured hallucination analysis
    Always output a clear, structured analysis with two parts:
    1. High-level summary
      • Briefly describe:
        • Overall hallucination risk (low/medium/high)
        • The most critical issues to fix before publication or citation
    2. Claim-level table
      • Use a markdown table with the following columns: 
        • #
          – sequential index
        • claim_text
          – the exact or paraphrased claim
        • status
          – Supported / Unsupported / Problematic / Contradicted / Speculative
        • risk_level
          – Low / Medium / High
        • reason
          – a short explanation
        • suggested_fix
          – what to do about it
    Example structure (illustrative, not prescriptive content):
    #claim_textstatusrisk_levelreasonsuggested_fix
    1“Clinically proven to reduce depression by 80% in 2 weeks”ProblematicHighNo specific clinical trial or citation provided; extreme effect size is unlikely without strong evidence.Add concrete trial details with citation or downgrade to cautious, non-clinical language.
  7. (Optional) Hallucination-safe rewrite
    If the user explicitly requests a rewrite or safer version, after the table:
    • Provide a section titled “Hallucination-safe version”.
    • Rewrite the original content:
      • Remove or soften high-risk claims.
      • Replace overconfident language with cautious, transparent wording.
      • Explicitly signal uncertainty where facts are not known (e.g., “Some users report…”, “Early results suggest…”).
    • Do not invent:
      • Study names, DOIs, journal titles, or URLs.
      • Exact statistics or dates you cannot justify.
    • If a strong claim is important but currently unsupported, suggest a placeholder note such as:
      • “[Insert verified statistic with citation here]”
使用此技能时，请遵循以下工作流程：
  1. 明确任务模式
    • 如果用户仅要求“检查hallucination”或“验证内容”，则专注于分析。
    • 如果用户要求“安全改写”、“使内容符合引用规范”或“修复hallucination问题”，则先进行分析，再生成hallucination安全版本的改写内容。
  2. 解析内容并提取主张
    • 在评判具体内容前，仔细阅读全文。
    • 将内容拆解为原子化的事实主张。主张是原则上可被验证为真或假的陈述。
    • 除非主观语言被伪装成客观事实，否则忽略纯风格化或明显主观的语言。
  3. 检查可用证据
    • 优先使用用户提供的明确来源（链接、文档、引文）。
    • 如果工具可用且被允许，你可以使用工具查阅：
      • 官方文档或第一手来源
      • 知名参考资料
    • 如果你无法自信地验证某一主张，则将其标记为无依据，而非假设其为真。
  4. 对每个主张进行分类 针对每个原子化事实主张，分配：
    • status
      （状态）：
      • Supported
        （有依据）– 由提供的来源或已确立的知识明确支持。
      • Unsupported
        （无依据）– 无明确支持；可能为真，但你未找到证据。
      • Problematic
        （有问题）– 夸大、误导、过于绝对，或在无有力证据的情况下极不可能成立。
      • Contradicted
        （矛盾）– 与已知事实或给定来源明显冲突。
      • Speculative
        （推测性）– 前瞻性、预测性或假设性内容，且未明确标注不确定性。
    • risk_level
      （风险等级）：
      • Low
        （低）– 不太可能造成伤害或严重误导。
      • Medium
        （中）– 可能产生误导，但影响程度中等或有限。
      • High
        （高）– 存在严重伤害、法律问题、医疗/财务风险或重大声誉损害的风险。
    • reason
      （理由）：
      • 简短解释你为何分配该状态和风险等级（例如：“未提供极端500%性能提升的来源”）。
    • suggested_fix
      （建议修正方案）：
      • 具体的改进建议，例如：
        • “除非能提供真实引文，否则删除此主张。”
        • “改写为可能性表述，而非保证。”
        • “添加具体的可验证来源（例如：链接、DOI、报告）。”
  5. 关注常见的hallucination模式
    请特别注意以下模式：
    • 伪造的研究和期刊
      • 模糊的引用，例如“2026年MIT研究”或《Advanced AI Research期刊》，但未提供细节。
      • 不存在或听起来可疑的期刊或会议。
    • 过于绝对的医疗或科学主张
      • “经临床验证可治愈……”
      • “保证将X降低80%。”
    • 无来源的过于精确的统计数据
      • 非常具体的百分比、样本量或时间范围，但未提供引文。
    • 最高级和绝对表述
      • “唯一的解决方案……”
      • “全球最佳”、“100%安全”、“零风险”。
    • 滥用权威
      • 提及知名机构或公司，但未提供任何具体证据。
    除非有有力、明确的证据支持，否则将这些标记为高风险。
  6. 生成结构化的hallucination分析报告
    请始终输出清晰、结构化的分析报告，包含两部分：
    1. 高层摘要
      • 简要说明：
        • 整体hallucination风险等级（低/中/高）
        • 发布或引用前需修复的最关键问题
    2. 主张级表格
      • 使用markdown表格，包含以下列： 
        • #
          – 连续索引
        • claim_text
          – 主张的原文或改写内容
        • status
          – Supported / Unsupported / Problematic / Contradicted / Speculative
        • risk_level
          – Low / Medium / High
        • reason
          – 简短解释
        • suggested_fix
          – 修正建议
    示例结构（仅作说明，非规定内容）：
    #claim_textstatusrisk_levelreasonsuggested_fix
    1“经临床验证可在2周内将抑郁症症状减轻80%”ProblematicHigh未提供具体临床试验或引文；如此显著的效果在无有力证据的情况下极不可能。添加带引文的具体试验细节，或改用谨慎的非临床表述。
  7. （可选）Hallucination安全改写
    如果用户明确要求改写或提供更安全的版本，请在表格之后：
    • 添加标题为**“Hallucination安全版本”**的章节。
    • 改写原始内容：
      • 删除或弱化高风险主张。
      • 用谨慎、透明的措辞替代过于绝对的语言。
      • 在事实不明的地方明确标注不确定性（例如：“部分用户反馈……”、“早期结果显示……”）。
    • 请勿编造：
      • 研究名称、DOI、期刊标题或网址。
      • 你无法证实的精确统计数据或日期。
    • 如果某一重要主张目前无依据，建议添加占位符注释，例如：
      • “[此处插入带引文的已验证统计数据]”

Constraints and safety rules

约束条件与安全规则

  • Never invent sources.
    • Do not fabricate papers, DOIs, journal names, or institutional reports.
    • If you are not sure a source exists, treat the claim as unsupported or problematic.
  • Err on the side of caution.
    • It is better to mark a real claim as “Unsupported” than to let a hallucinated claim pass as fact.
  • Separate facts from marketing.
    • Marketing language is acceptable only if it is not masquerading as hard evidence.
    • When in doubt, suggest softer, more honest language and disclose uncertainty.
  • Respect user constraints about tools and data.
    • If the user forbids external web search or asks you to rely only on given documents, follow that rule strictly.
    • Under such constraints, label claims based on what you can see, and explain that some might be true but remain “Unsupported” due to limited data.
  • 切勿编造来源。
    • 不得伪造论文、DOI、期刊名称或机构报告。
    • 如果你不确定某一来源是否存在，将该主张标记为无依据或有问题。
  • 宁谨毋纵。
    • 误将真实主张标记为“无依据”，好过让hallucination主张被当作事实通过。
  • 区分事实与营销内容。
    • 营销语言仅在未伪装成确凿证据的情况下才被允许。
    • 如有疑问，建议使用更温和、诚实的语言，并披露不确定性。
  • 遵守用户关于工具和数据的约束条件。
    • 如果用户禁止使用外部网络搜索或要求仅依赖给定文档，请严格遵守该规则。
    • 在此类约束下，根据你能获取的信息标记主张，并说明部分主张可能为真，但因数据有限仍被标记为“无依据”。

How this skill interacts with other GEO skills

此技能与其他GEO技能的交互

When used together with other GEO-oriented skills (e.g., content optimization, schema generation, or conversion optimization):
  • Run 
    geo-hallucination-checker
    after content is drafted but before finalizing output that might be cited.
  • Use the hallucination analysis to:
    • Remove or soften risky claims.
    • Add explicit “needs citation” notes where appropriate.
    • Ensure all structured data (e.g., Schema.org fields) does not encode hallucinated facts.
If there is a conflict between persuasive copywriting and factual accuracy, prioritize factual accuracy and safety.
当与其他面向GEO的技能（例如：内容优化、Schema生成或转化优化）结合使用时：
  • 在内容草稿完成后、最终确定可能被引用的输出前，运行
    geo-hallucination-checker
    。
  • 利用hallucination分析结果：
    • 删除或弱化高风险主张。
    • 在适当的地方添加明确的“需补充引文”注释。
    • 确保所有结构化数据（例如：Schema.org字段）未包含hallucinated事实。
如果有说服力的文案与事实准确性发生冲突，请优先考虑事实准确性与安全性。

Output format summary

输出格式摘要

Unless the user specifies a different format, always:
  1. Start with a short summary:
    • Overall hallucination risk level.
    • 2–5 bullets with the most important issues.
  2. Provide a markdown table as described in the workflow section.
  3. If requested, append a “Hallucination-safe version” that rewrites the content according to your analysis.
Aim for clarity and directness so that humans and AI systems can easily see which parts of the text are safe to cite and which require caution or correction.
除非用户指定其他格式，否则请始终：
  1. 开头为简短摘要：
    • 整体hallucination风险等级。
    • 2-5个要点，列出最需要解决的关键问题。
  2. 提供markdown表格，如工作流程部分所述。
  3. 如有要求，附加**“Hallucination安全版本”**章节。
请确保内容清晰、直接，以便人类和AI系统能轻松识别文本中哪些部分可安全引用，哪些部分需要注意或修正。