AI Methodology

How Visibility Engine Generates Guidance

1. Collect first-party signals from connected data sources.
2. Consolidate and normalize metrics into shared analytics contracts.
3. Apply deterministic detectors (trend, anomaly, intent, health checks).
4. Use AI for explanation, prioritization, and rewrite/forecast assistance.
5. Track recommendation outcomes and feed learning context back into prompts.

Why This Is Not Generic LLM Output

• Action-tracking loop with impact measurement rather than static advice.
• Cross-signal diagnosis (traffic, intent, page quality, and trendline context).
• Tier-aware outputs with explicit premium/credit boundaries.

Prompt and Logic Surfaces

These source modules are versioned in the repository and updated through standard code review workflows.

• src/ai/prompt-builder.mjs
• src/ai/recommendation-engine-v2.mjs
• src/feedback/learning-context.mjs

Representative Output Types

• Priority recommendation with projected impact and affected pages.
• AI briefing with situation, one-thing focus, risks, and actions.
• Content rewrite suggestion tied to observed CTR/position gaps.