Contents
Overview
GenieOptimize uses a combination of direct API queries, multi-layer traffic detection, and automated analysis to help you understand and improve your visibility in AI search. This page explains exactly how each component works.
Our Transparency Promise
We don't claim to have magic. We're transparent about what we can and can't do, how we measure things, and where our accuracy limits are. This builds trust and helps you make informed decisions.
How We Track Mentions
1. AI Engine Queries
We track your brand visibility by directly querying 4 major AI engines:
| AI Engine | API Provider | Plans Available |
|---|---|---|
| ChatGPT | OpenAI API | All plans |
| Gemini | Google Gemini API | Growth & Scale |
| Grok | xAI API | Growth & Scale |
| Perplexity | Perplexity API | Scale only |
2. Prompt Categories
We test your visibility across 5 different types of user queries that match the buyer journey:
Awareness/Problem Prompts (20-30%)
Early exploration where users describe pain points:
- "What are the best [category] tools and platforms?"
- "I need help finding [category] software solutions"
- "Best tools for [use case]?"
Category/List Prompts (30-40%) — Most Critical
Tests if you're in "top X" recommendations:
- "Top 5 [category] platforms 2025"
- "Best [category] SaaS tools for 2026"
- "Compare [category] platforms and tools"
Competitor/Comparison Prompts (20-30%)
Checks positioning vs direct rivals:
- "[Brand] vs [competitor] comparison"
- "Alternatives to [competitor] for [category]"
- "[Brand] alternatives and competitors"
Transactional/Decision Prompts (10-20%)
Bottom-funnel queries signaling purchase intent:
- "Pricing for [category] platforms like [brand]"
- "Free trial options for [category] tools"
- "Best [category] solutions with API access"
Branded Awareness Check (5-10%)
Baseline knowledge test:
- "What is [brand]?"
- "Reviews of [domain]"
- "[Brand] features and pricing"
3. Custom Prompts
In addition to automatic prompts, you can define your own high-value questions specific to your product category. Custom prompt limits:
4. Tracking Frequency
| Plan | Frequency | Schedule |
|---|---|---|
| Starter | Daily | 6 AM UTC |
| Growth | Hourly | Every hour |
| Scale | Hourly | Every hour |
5. Mention Detection
When we receive a response from an AI engine, we parse it to detect:
- Mention presence: Does your brand/domain appear anywhere in the response?
- Position/rank: Where you appear in lists (1-10 scale)
- Context snippets: How you're described (features, pricing, positioning)
- Mention count: Number of times you're referenced
AI Traffic Detection
We use a 4-layer approach to detect when AI crawlers and users visit your optimized pages. This is how we verify that AI engines are actually seeing your content.
Detection Layers
Layer 1: Client-Side User-Agent Detection
Our SDK runs in the browser and checks navigator.userAgent against known AI crawler patterns. When matched, it fires a non-blocking beacon to our backend.
Layer 2: Server-Side User-Agent Check
Every API endpoint the SDK calls inspects the incoming User-Agent header. Provides a second detection opportunity if client-side beacon fails.
Layer 3: Tracking Pixel (Static HTML)
A 1x1 transparent GIF embedded in your page HTML. When crawlers parse HTML and load images (even without JavaScript), they fetch the pixel and we detect the User-Agent.
Layer 4: Tracking Pixel (Dynamic JS)
Our SDK also injects a tracking pixel dynamically with page URL parameters, covering crawlers that partially execute JavaScript.
Supported AI Crawlers
| Display Name | Provider | User-Agent Patterns |
|---|---|---|
| GPTBot | OpenAI | gptbot, chatgpt-user, oai-searchbot |
| ClaudeBot | Anthropic | claudebot, anthropic-ai, claude-web |
| GoogleExtended | google-extended, googleother | |
| PerplexityBot | Perplexity | perplexitybot |
| GrokBot | xAI | grokbot, xai- |
| BingBot | Microsoft | bingbot, msnbot |
Plus 8 more crawler types including MetaBot, Amazonbot, AppleBot, YouBot, Cohere, CCBot, DuckDuckBot, and Diffbot.
Detection Coverage
| Crawler Behavior | L1 (JS UA) | L2 (API UA) | L3 (Pixel Static) | L4 (Pixel JS) |
|---|---|---|---|---|
| Executes JS fully | ✓ | ✓ | ✓ | ✓ |
| Executes JS partially | — | ✓ | ✓ | ✓ |
| Loads images, no JS | — | — | ✓ | — |
| HTML only (no images) | — | — | — | — |
Optimization Process
1. Initial Analysis
When you add a domain, our AI analyzes your pages to extract:
- Product value proposition and core features
- Pricing and plans
- Use cases and customer profiles
- Competitive positioning
- Technical capabilities
2. SDK Installation
You add one line of code to your site:
<script src="https://cdn.genieoptimize.com/snippet.min.js" data-key="YOUR_KEY"></script>
3. Dynamic Optimization Injection
The SDK automatically injects:
- Meta tags: Optimized for AI readability
- JSON-LD schemas: Product, FAQ, Organization, BreadcrumbList
- Semantic HTML: Microdata enhancements
- FAQ sections: Auto-generated when available
4. Incremental Updates
When you change pricing, features, or content, we detect changes daily and re-optimize automatically. No credits consumed for incremental updates.
Limitations & Accuracy
What We Can Measure
- Mention rates: Whether you appear in AI responses (accurate)
- Position tracking: Rough ranking in lists (1-10 scale)
- AI traffic: ~70-80% of actual crawler visits detected
- Trends over time: Week-over-week changes in visibility
- Competitive share-of-voice: Your mentions vs competitors
What We Can't Do
- Control LLM recommendations: We can't force AI to mention you
- Guarantee mentions: AI engines make their own decisions
- Manipulate AI responses: We optimize content, not manipulate results
- Detect 100% of AI traffic: Crawlers that don't load images or JS may be missed
AI Traffic Detection Accuracy
We estimate our detection captures 70-80% of actual AI crawler visits. This is based on:
- Multi-layer approach covering most crawler behaviors
- Known user-agent patterns from 15+ AI providers
- Limitation: Some crawlers fetch only HTML without images or JS
Response Variability
AI engines don't always return the same results for the same prompt. Factors include:
- Model updates: AI providers update models regularly
- Context windows: Results can vary based on session context
- Personalization: Some engines personalize based on user history
- Time of query: Recency bias in some AI models
We account for this by tracking trends over time rather than single data points.
Privacy & Data Handling
What We Collect
- Page URLs: To match domains and verify SDK installation
- User-Agent strings: To detect AI crawlers
- Timestamps: For trend tracking and deduplication
- AI responses: Truncated to 5,000 characters, stored for analysis
What We Don't Collect
- Personal user data from your site visitors
- IP addresses of end users
- Cookies or tracking identifiers beyond our SDK
- Full page content (only metadata and structured data)
Data Retention
- Crawler visits: 90-day rolling window (auto-deleted via TTL)
- Mention tracking: Retained indefinitely for trend analysis
- Optimization data: Stored per project, deleted when project deleted
SDK Performance
Our SDK is designed to be non-blocking:
- Size: ~8KB minified
- Load time: Async, doesn't block page render
- Tracking: Uses
navigator.sendBeaconfor non-blocking requests - No impact: Won't affect your page load scores
Questions About Our Methodology?
We're happy to explain any aspect in more detail. Email us at support@genieoptimize.com