GEO Sniffer

Description

Reveals how AI models (ChatGPT, Claude, or Gemini) interpret any topic — what search terms they generate, which websites they trust, and which brands they recommend. Runs four analysis methods in parallel and produces a single unified report with actionable GEO (Generative Engine Optimization) insights.

Core Capabilities

• Uncovers the AI's internal research process: the exact search strings it would generate when answering a query
• Maps the AI's source trust network: which websites it considers authoritative for a given topic
• Reveals brand shelf positioning: which brands the AI recommends, in what order, and with what category labels (e.g., "premium", "budget", "luxury cost-performance")
• Detects embedded brand-topic associations: brands that appear inside search strings before the AI even starts researching
• Cross-language market analysis: discover how the same topic is framed differently across language markets (e.g., BYD is "budget" in English but "luxury cost-performance" in Japanese)
• Pattern analysis: identifies search term clustering, ranking format preferences, and source diversity signals

Good Query Examples

• Query: "best running shoes 2025" | Model: ChatGPT → Discover which brands ChatGPT puts on the shelf, in what order, and with what labels — does it call Hoka "premium" or "performance"? Which review sites does it trust: Runner's World, Wirecutter, or niche running blogs?
• Query: "Ducati vs BMW motorcycles" | Model: ChatGPT → See how the AI frames a head-to-head comparison: which brand it positions as the leader, what attributes it assigns to each (performance vs. touring), and whether its search strings reveal a bias toward one brand
• Query: "best luxury EV" | Model: ChatGPT | Target Language: Japanese → Compare how the same topic is framed for different markets — a brand labeled "budget" in English might appear as "luxury cost-performance" in the Japanese response. Report stays in English with translation columns

Bad Query Examples

• "Analyze how ChatGPT, Claude, and Gemini all respond to 'best project management tools'" → This agent analyzes one AI model per run. Fix: submit three separate jobs — one for each model — and compare the reports.
• "Does bmw.com get mentioned when people ask ChatGPT about luxury cars?" → This asks about a specific domain's visibility in AI responses — that's what SEO & GEO Researcher does (domain-first measurement). GEO Sniffer is topic-first discovery. Fix: query "best luxury cars" and look for BMW in the results, or use SEO & GEO Researcher for domain mention tracking.
• "Which brands does ChatGPT recommend for running shoes?" → Don't phrase it as a meta-question about AI. Write the query as a normal user would type it. Fix: "best running shoes 2025" — the report itself reveals the brands.
• "marketing" → Too vague. The AI will generate generic search strings that aren't useful for any specific brand or category. Fix: "B2B content marketing tools" or "social media marketing for restaurants."
• "best electric vehicles, also check keyword rankings and backlink profile for tesla.com" → GEO Sniffer only analyzes AI model behavior. It does not provide SEO data (rankings, backlinks, search volume). Fix: use GEO Sniffer for the AI analysis and SEO & GEO Researcher for the SEO data.

Use-Case Ideas

• Brand audit — check whether your brand appears in AI recommendations for your category, and how it's positioned vs. competitors
• Content strategy — discover the exact search terms AI models generate for your topic, then create content that matches those patterns
• Source mapping — identify which websites the AI trusts for your category, then prioritize those for PR and editorial placements
• Brand positioning intelligence — learn how AI categorizes your brand (e.g., "premium" vs. "value") and compare across models
• Cross-market discovery — see how the same topic is framed in different language markets to inform localization strategy
• AI search term seeding — use AI-generated search strings as keyword seeds for SEO and paid search planning
• Competitive shelf analysis — understand the full "AI shelf" for a category: who's on it, in what order, and with what labels

Limitations

• Analyzes one AI model per run — to compare ChatGPT, Claude, and Gemini, run three separate jobs
• Topic-first, not domain-first — does not measure whether a specific domain is cited in AI responses (use SEO & GEO Researcher for that)
• No location parameter — the API responds based on query language only; "German in Germany" and "German in Switzerland" produce identical results
• Cross-language analysis relies on query translation — nuance may vary slightly from a native speaker's natural phrasing
• The live search method occasionally includes English search strings even for non-English queries (cross-language leakage)
• One query per analysis, up to 500 characters

Input/Output Spec

Input:

• query (required): The topic to analyze — phrased as a normal user would ask it, not as a meta-question about AI. 3-500 characters.
• model (required): Which AI model to analyze. Options: chatgpt (OpenAI), claude (Anthropic), gemini (Google).
• target_language (optional): Set this when your query is in one language but you want to analyze a different language market. The query will be translated to the target language before analysis, and the report stays in the query's language with translation columns. Leave blank to analyze in the query's own language. Options: English, Japanese, Chinese, Korean, German, French, Spanish, Portuguese, Russian, Arabic, Italian, Dutch, Hindi, Turkish, Thai, Polish, Swedish, Indonesian, Vietnamese, Persian, Czech, Romanian, Ukrainian.

Output:

• Markdown report with:
• Executive summary with key takeaways
• Search terms: keyword-style and natural-language searches the AI generates, with pattern analysis
• Trusted sources: websites the AI considers authoritative, with landscape analysis
• Brand recommendations: brands with category labels, positioning analysis, and competitive ordering
• Connecting the Dots: cross-section insights that emerge from all data together
• Actionable next steps
• When cross-language: translation columns in all data tables, translated query shown in header

Disambiguation Signals

This agent reveals how AI models internally interpret topics — search terms, trusted sources, and brand recommendations. It is topic-first and qualitative: "How does the AI think about luxury EVs?" It does not measure domain visibility or provide SEO ranking data.

• GEO Sniffer vs. SEO & GEO Researcher: This is the most important distinction. Both deal with AI visibility, but they answer fundamentally different questions:
• GEO Sniffer = "How does the AI interpret this topic?" → topic-first, qualitative discovery. Input is a topic/query. Output is search strings, trusted sources, brand recommendations with category labels and positioning.
• SEO & GEO Researcher = "Does my domain appear in AI responses?" → domain-first, quantitative measurement. Input is a domain. Output is mention counts, citation frequency, competitive domain rankings, and can also scrape actual ChatGPT responses. Also provides traditional SEO data (keyword rankings, traffic, backlinks).
• Route to SEO & GEO Researcher when: the user asks about a specific domain's presence in AI responses (e.g., "Is bmw.com mentioned in ChatGPT?"), wants AI mention metrics and scores, or needs any traditional SEO data alongside AI visibility.
• Route to GEO Sniffer when: the user wants to understand how an AI model frames a topic, which brands it recommends, or what sources it trusts — without starting from a specific domain.
• Route to Advanced Web Research when: the user wants broad multi-source web research, not specifically AI model behavior (e.g., "Research EV adoption trends in Europe")
• Route to News Research when: the user wants recent news coverage and media sentiment (e.g., "What's been in the news about Tesla this week?")
• Route to GWI Spark when: the user wants real consumer survey data about audience demographics or behavior (e.g., "What do millennials think about electric vehicles?")
• Route to Statista when: the user wants market statistics, revenue data, or industry forecasts (e.g., "What is the global EV market size?")