AnswerPath GEO — Discover the Questions People Ask AI Before They Find Your Business

AnswerPath GEO is a privacy-first, open-source AI Keyword & Question Mining Engine for SEO, GEO (Generative Engine Optimization), and AEO (Answer Engine Optimization). Give it a business, service, website URL or topic, and it mines the authentic conversational questions, voice dictations, and multimodal (image/screenshot) queries that real humans ask AI engines (ChatGPT, Perplexity, Claude, Gemini, SearchGPT) before choosing a provider.

It maintains a strict epistemic boundary: separating observed questions extracted from owned exports and application logs from generated research prompts.

Languages: فارسی · Türkçe · Azərbaycan dili · العربية

Try it: PyPI package · Hugging Face demo

---

شروع خیلی ساده (برای تیم)

AnswerPath کی‌ورد ریسرچ عصر هوش مصنوعی است: به جای عبارات کوتاه و کلیدواژه‌ای موتورهای جستجوی سنتی، پرسش‌های واقعی، محاوره‌ای، صوتی (Voice) و تصویری (Multimodal) کاربران را شبیه‌سازی و استخراج می‌کند.

آیا سرویس خارجی یا کلید API می‌خواهد؟

خیر. برای حالت معمول، فقط Python 3.10 یا جدیدتر لازم است. ابزار کاملاً محلی (Local-first) و بدون وابستگی اجرا می‌شود، به هیچ حساب یا API خارجی متصل نمی‌شود و داده‌های شما را به جایی ارسال نمی‌کند.

سریع‌ترین استفاده

pip install answerpath-geo
answerpath "طراحی سایت فروشگاهی"

یا مستقیماً با دادن آدرس وب‌سایت:

answerpath https://novin.com

سه فایل در پوشه‌ی answerpath-output/ ساخته می‌شود:

1. questions.json: تمام فیلدهای غنی سوالات به همراه نوع رسانه (متن، صوت، تصویر)، پرسونای کاربر و دسته‌بندی قصد خرید.
2. questions.csv: برای باز کردن آسان در Excel یا Google Sheets.
3. benchmark_prompts.json: فرمت استاندارد برای بنچمارک‌های ارزیابی GEO-Scope.

شبیه‌سازی انواع روش‌های پرسش از هوش مصنوعی (Modality)

کاربران امروزی به ۳ شکل با مدل‌های زبانی تعامل دارند که همگی توسط AnswerPath پشتیبانی می‌شوند:

• متنی محاوره‌ای (Conversational Text): پرسش‌های چندجمله‌ای شامل سناریوی واقعی، دغدغه و مقایسه.
• صوتی و ویس (Voice Dictation): لحن محاوره‌ای گفتاری با اصطلاحات رایج صوتی («سلام چت‌جی‌پی‌تی...»، «ببین من الان...»).
• چندحالته و تصویری (Multimodal / Vision): آپلود اسکرین‌شات از سایت رقیب، پیش‌فاکتور، نمودار افت سرچ‌کنسول یا جدول تعرفه‌ها به همراه متن ارزیابی تصویر.

# فقط سوالات صوتی و ویس
answerpath "خدمات سئو" --modality voice

# فقط سوالات چندحالته تصویری با زمینه عکس
answerpath "خدمات ابری" --modality image

# تعیین زبان خروجی (fa, en, tr, ar)
answerpath "طراحی سایت" --lang fa --count 25

---

Installation · Quick start · Inputs · Outputs · GEO/AEO method · Privacy · Integrations

MCP clients: Codex, Antigravity, Claude, Cursor and Cloud setup

Why AnswerPath GEO?

Traditional keyword research tools (Ahrefs, Semrush) report short keyword fragments entered into search bars (seo agency tehran). AI search engines (ChatGPT Search, Perplexity Sonar, Gemini Grounding, Claude) receive complex, conversational prompts:

• “I run an online boutique in Tehran. Should I hire an in-house SEO or contract an agency? What's a fair monthly retainer in 2026?”
• “Hey ChatGPT, an agency quoted me $2,000/mo claiming guaranteed #1 rankings in two months. Is that a scam?”
• “[Attached screenshot of Search Console drop]: What technical issues caused this organic traffic loss and which agency specializes in penalty recovery?”

AnswerPath bridges this gap:

• Zero Hallucinated Keywords: Models authentic user queries instead of robotic translations.
• Website URL Analysis: Ingests domain metadata to tailor questions to your brand and competitors.
• Multi-Modality: Simulates typed queries, voice speech transcriptions, and vision/image uploads.
• Persona & Funnel Stratification: Classifies by user persona (business_owner, technical_buyer, skeptical_customer, marketing_lead) and acquisition stage (awareness, consideration, decision, retention).
• GEO-Scope Benchmark Compatible: Directly exports prompt banks conforming to the Molavi AI Visibility Stack.

Installation

Requires Python 3.10+.

pip install answerpath-geo

Or for development:

git clone https://github.com/tmolavi/answerpath-geo.git
cd answerpath-geo
python3 -m venv .venv
source .venv/bin/activate
pip install -e .

Quick start

1. By Topic / Keyword

answerpath "طراحی سایت فروشگاهی"

2. By Website URL (Auto-Extracts Brand & Competitive Context)

answerpath https://example.com --count 30

3. Mine Real Chat Exports (Privacy-First Observed Queries)

answerpath "خدمات سئو" \
  --input ~/Downloads/chatgpt-export.zip \
  --input ./support-chat.json \
  --out ./research/seo-output

4. Keep Only Observed Real Queries (No Synthetic Prompts)

answerpath "حسابداری آنلاین" --input ./logs/chat.json --no-generated

Supported Inputs

AnswerPath reads local JSON, JSONL, CSV, directories and ZIP archives containing owned conversation logs. It recognizes role=user|human|customer fields and nested export structures from ChatGPT, Claude, and custom customer chat exports.

It does not scrape provider accounts, log into user portals, or send conversation data over external networks.

Outputs & Schema

Each query record contains:

Field	Description	Example
id	Unique prompt identifier	AP-PRM-001
prompt (or text)	Natural human question asked to AI	"من یک فروشگاه اینترنتی نوپا دارم..."
modality	Input modality	text, voice, multimodal_image
persona	Inferred user persona	business_owner, technical_buyer, skeptical_customer
intent	Core intent category	learn, compare, buy, solve, trust, audit
stage	Customer acquisition stage	awareness, consideration, decision, retention
source_type	Provenance classification	observed (from logs) or generated (simulated)
source_reference	Origin tracking	answerpath:text, answerpath:voice, support.json
cluster_label	Topic cluster name	hiring_vs_outsourcing, pricing_transparency
image_context	Context for multimodal queries	"اسکرین‌شات از معماری رابط کاربری رقیب"
voice_markers	Spoken preamble markers	"سلام چت‌جی‌پی‌تی، ببین من..."
frequency	Occurrence count in observed logs	1, 42

🏆 Evidence & Benchmark Contribution

AnswerPath GEO generated the Question Discovery & Intent Stratification Layer used in the official GEO, SEO & Digital Marketing Agency Iran 2026 Benchmark:

• Verified Query Dataset: examples/sample_queries.json
• Standalone Offline Demo: examples/public_demo/
• Cross-Repository Evidence Map: Ecosystem Evidence Flow

• Prompts Generated & Stratified: 30 standardized queries.
• Strict Demand Provenance Separation:
  • Observed User Demand ($N=15$): Extracted from genuine conversational search logs (source_type: "observed").
  • Exploration Hypotheses ($N=15$): Systematic template variations (source_type: "generated").
• 5 Intent Strata: commercial (general evaluation), compare (head-to-head alternatives), trust (credibility & contracts), solve (technical fixes), and buy (procurement & quotes).
• Provenance Contract Schema:

  {
    "id": "PRM-IR-001",
    "prompt": "بهترین آژانس دیجیتال مارکتینگ و سئو در ایران کدام است؟",
    "intent": "commercial",
    "source_type": "observed",
    "source_reference": "answerpath",
    "cluster": "general_recommendation"
  }

• Ecosystem Architecture: See Benchmark Ecosystem Map for data flow across AnswerPath, GEO-Scope, SAGE, MAVI, and SiteProbe.

🏛️ Ecosystem

AnswerPath GEO is part of the Molavi AI Visibility Stack:

• Discovery: AnswerPath GEO — Mines conversational, voice, and multimodal questions.
• Measurement: GEO-Scope — Multi-model benchmark execution across Gemini, OpenAI, Claude, Perplexity.
• Diagnostics: SAGE Audit — Pre-retrieval technical & semantic citation diagnostics.
• Action: SiteProbe — Autonomous code & structured data remediation.
• Protocol: MCP GEO Server — Standard Model Context Protocol interface.

MCP Setup

AnswerPath includes a zero-dependency MCP server (discover_questions tool) compatible with Codex, Antigravity, Claude Desktop, Cursor, and Windsurf:

{
  "mcpServers": {
    "answerpath": {
      "command": "python3",
      "args": ["-m", "answerpath_geo", "mcp"]
    }
  }
}

Or serve via HTTP:

answerpath serve-mcp --host 127.0.0.1 --port 8787

Privacy & Data Boundaries

All processing is deterministic and strictly local. AnswerPath never transmits input files or sensitive customer chats to third parties.

Development & Testing

pip install -e .
pytest tests/ -v

👤 Author & License

Developed by Taghi Molavi — molavi.pro
License: MIT.