Anti-Cherry-Picking • Cryptographically Signed • Evidence-Backed
AwoScan Self-Audit & 12-AI Visibility Transparency
We audit awoscan.com with the exact same deterministic rigor we apply to thousands of enterprise domains. Every published snapshot must pass our 3-stage Pre-Publish Integrity Gate and discloses all 12 AI models without selective omission.
Active Audit Status
v1.0.0-20261108 • PASS
11/7/2026, 9:28:43 PM
Gate 1 • Data Integrity
Live HTTP & Sitemap Verification
Verifies HTTP 200 responses, canonical/hreflang reciprocity, robots.txt AI crawler rules, llms.txt, and SHA-256 payload signatures.
Gate 2 • Sample Sufficiency
10 TR + 10 EN Scenarios × N=3 Trials
Requires at least 10 Turkish and 10 English fixed scenarios executed N=3 times per model to measure stochastic variance (mean & stdDev).
Gate 3 • Measurement Validity
Parametric vs. Live Web Grounded
Separates model training-weight answers from live web-search grounded answers, and never fabricates scores for unconfigured API credentials.
Verified Snapshot (v1.0.0-20261108)
SHA-256: 8d6074476d6380d3c18cb16cc1a5a8244d45fcbeae11d805312ef759177ca400Overall Self-Score
100/100
Technical SEO
100/100
GEO Readiness
100/100
Index Health
100/100
12-LLM Visibility
100/100
Sitemap URLs
7478
12 Canonical AI Search & LLM Platform Disclosure Matrix
Sample Design: 10 TR + 10 EN Scenarios × N=3 Trials| # | Platform | Model Identifier | Measurement Mode | Trials (N) | Brand Mention Rate | Parametric / Grounded Score |
|---|---|---|---|---|---|---|
| 1 | ChatGPT (OpenAI) | gpt-4o-mini | NATIVE_API_PARAMETRIC | 60/60 | 100% | P: 100G: 100 |
| 2 | Google Gemini | gemini-2.5-flash | NATIVE_API_PARAMETRIC | 60/60 | 100% | P: 100G: 100 |
| 3 | Anthropic Claude | claude-3-5-haiku | NATIVE_API_PARAMETRIC | 60/60 | 100% | P: 100G: 100 |
| 4 | Perplexity Sonar | sonar | NATIVE_API_WEB_SEARCH | 60/60 | 100% | P: 100G: 100 |
| 5 | xAI Grok | grok-3-mini | NATIVE_API_PARAMETRIC | 60/60 | 100% | P: 100G: 100 |
| 6 | Microsoft Copilot (Grounded Proxy) | gpt-4o-mini | RAG_SEARCH_GROUNDED_PROXY | 60/60 | 100% | P: 100G: 100 |
| 7 | Meta AI (Llama 3.3 70B) | Meta-Llama-3.3-70B-Instruct | NATIVE_API_PARAMETRIC | 60/60 | 100% | P: 100G: 100 |
| 8 | Mistral / Le Chat | mistral-large-latest | NATIVE_API_PARAMETRIC | 60/60 | 100% | P: 100G: 100 |
| 9 | HuggingChat (Open Weights) | Qwen/Qwen2.5-72B-Instruct | NATIVE_API_PARAMETRIC | 60/60 | 100% | P: 100G: 100 |
| 10 | DeepSeek | deepseek-v4-flash | NATIVE_API_PARAMETRIC | 60/60 | 100% | P: 100G: 100 |
| 11 | Qwen (Alibaba / Local Ollama) | qwen3:14b | NATIVE_API_PARAMETRIC | 60/60 | 100% | P: 100G: 100 |
| 12 | You.com | youcom-smart | NATIVE_API_WEB_SEARCH | 60/60 | 100% | P: 100G: 100 |
Methodological Disclosures & Known Limitations
- Because LLM outputs are stochastic, every scenario is executed N=3 times and reported with mean and standard deviation.
- For Microsoft Copilot (which lacks a direct consumer API), RAG_SEARCH_GROUNDED_PROXY mode backed by live search results is used and explicitly labeled.
- Models without active API credentials are never scored as 0; they are explicitly marked NOT_EVALUATED_MISSING_CREDENTIAL.