How we measure AI answer-engine visibility.
The full methodology page is in draft. Today's free check runs four static scanners — the AI-crawler robots.txt audit (14 user-agents, explicit/fall-through distinction), the llms.txt validator (with honest framing about who actually respects it), the schema.org JSON-LD extractor (with FAQPage flagged as the highest-leverage gap per Relixir 2025), and the HTTP-header inspector. Live engine probes across ChatGPT, Claude, Gemini, and Perplexity arrive in Stage 2.
When the Audit tier ships, this page expands to cover the full measurement design: stratified prompt sets across four intent clusters, N=10 sampling per cell pooled to intent-cluster × model × region, bootstrap confidence intervals with 2,000 resamples, model snapshotting via nightly control prompts plus vendor-changelog scraping, and the URL-hallucination exclusion policy. Every academic reference will live here too:
- Sielinski, R. (2026). Quantifying Uncertainty in AI Visibility: A Statistical Framework for Generative Search Measurement. arxiv:2603.08924
- Schulte, J., Bleeker, M., Kaufmann, P. (2026). Don't Measure Once: Measuring Visibility in AI Search (GEO). arxiv:2604.07585
- Sharma, A. P. (2026). The Discovery Gap: How Product Hunt Startups Vanish in LLM Organic Discovery Queries. arxiv:2601.00912
- Aggarwal, P. et al. (2024). GEO: Generative Engine Optimization. arxiv:2311.09735
Questions, corrections, or critiques: methodology@canaifind.com. We log them publicly with disposition.