SiteNexis Scout: Intent Classification as a Cognitive Layer
Scout v1 adds a stateless cognition layer to SiteNexis. The Intent Engine classifies every page by query intent, scores coverage distribution, and identifies structural mismatches between what a page claims to do and how it is structured.
Traditional content analysis asks: is this page technically sound? Scout asks: what is this page trying to accomplish, and does its structure support that goal? The Intent Engine classifies each page into one of seven intent types and then scores whether the page's heading structure, schema types, URL patterns, and content signals are aligned with that intent. A mismatch between stated purpose and structural signals is an invisible visibility problem — the page looks correct to humans but sends confused signals to AI retrieval systems.
The Seven Intent Types
- Informational — the page answers a factual question or explains a concept
- Commercial — the page compares options, reviews products, or supports a purchase decision
- Navigational — the page directs users to a specific destination or section
- Research — the page provides data, methodology, or in-depth analysis for a specialist audience
- Creation — the page supports content creation: templates, generators, tools
- Learn and Solve — the page teaches a skill or solves a specific problem step-by-step
- Local — the page serves a geographically defined user with location-specific content
How Classification Works
Scout sends each page's title, H1, meta description, and first 500 words to Qwen 2.5 via OpenRouter with a structured classification prompt. The model returns an intent type plus confidence score. If the Claude API call fails or returns a low-confidence result, Scout falls back to a keyword regex classifier — a deterministic rule set matching URL patterns, heading structures, and common intent signals. The fallback ensures classification always completes even when the AI API is unavailable.
Intent Alignment Scoring
After classification, Scout scores alignment between the assigned intent and the page's structural signals. Informational pages are checked for FAQ schema, definition structures, and H2 headings framed as questions. Commercial pages are checked for Product or Review schema, comparison tables, and rating structures. Learn and Solve pages are checked for HowTo schema, numbered steps, and procedure headings. Each signal check produces a 0–1 score. The alignment composite identifies pages where intent and structure diverge.
▲A common pattern we detect: pages classified as Commercial but missing Product schema and comparison structures — they are optimized for human browsing but not for AI commercial query retrieval. Structural alignment with intent is as important as content quality.
Coverage Distribution Analysis
At the domain level, Scout computes Shannon entropy across the intent distribution. A domain where 90% of pages are classified as Informational has low intent entropy — it may have gaps in Commercial or Learn and Solve coverage that represent unaddressed query surfaces. The intent distribution dashboard shows coverage balance and flags intent types with no pages or thin coverage, enabling strategic content gap planning.
Classify every page on your domain by query intent.
Run Scout analysis