METHOD 1.0.0 · CHECK SET 2026-08-30

Topic Alignment Lab methodology

An editorial coverage model that asks whether a page resolves the important parts of a stated user need. It is not keyword density, a ranking score or proof of search demand.

What “topic alignment” means

A page is aligned when its useful, extractable passages collectively address the distinct subtopics—or facets—needed to satisfy a target question for the intended audience. Repeating a phrase does not count as deep coverage. A concise passage can cover a facet strongly when it directly explains the answer, boundary, step or distinction.

Analysis process

  1. Fetch one public HTML page with response-size, redirect and time limits.
  2. Prefer main or article content and remove scripts, styles, navigation, forms and footer regions.
  3. Split content into heading-bounded passages with a minimum substantive length.
  4. Ask a disclosed small language model to infer user intent and six to nine distinct expected facets.
  5. Score each facet for relevance and completeness, weighted by its 1–5 importance.
  6. Attach the strongest matching passage, explain the gap and propose one editorial move.
  7. If model analysis fails, use a clearly labelled deterministic term-coverage fallback with low confidence.

Scoring

overall = Σ(facet coverage × facet importance) / Σ(facet importance)

Facet states are Covered at 75–100, Partial at 35–74 and Gap at 0–34. Labels are descriptive rather than industry benchmarks: 85–100 Deeply aligned; 65–84 Aligned, with gaps; 45–64 Partial topic coverage; below 45 Weak topic alignment.

The score does not include backlinks, technical SEO, citations, evidence quality, conversion performance or observed rankings. Those are separate questions and separate tools.

Evidence and uncertainty

Model-assisted estimate

The model receives the user’s topic, optional audience and capped passage excerpts. It is told to assess answer relevance and completeness rather than repetition. The report stores the exact model ID.

Passage evidence

Every supported facet should name one extracted passage. Evidence is a short paraphrase; the full passage excerpt remains visible in the report. A match does not validate factual accuracy.

Deterministic fallback

If the semantic call fails or returns unusable structure, distinctive target terms become provisional facets and exact passage occurrences determine a low-confidence coverage estimate. This fallback is deliberately conservative.

Editorial suggestions

Recommendations describe missing informational work. They should be reviewed against real expertise, source evidence, brand position and the page’s intended scope before publication.

Limits

  • The target topic is user supplied; the tool does not estimate search volume or commercial value.
  • Only one page is analysed, so another URL may intentionally own a missing facet.
  • Raw response HTML is inspected; interaction-only content may be absent.
  • The model can misclassify intent, merge facets or overestimate implied coverage.
  • The tool does not fact-check claims or assess source quality.
  • English-language pages and questions are best tested.
  • Results can vary after model changes; stored reports remain frozen with their version and model ID.
  • Use the map to form an editorial hypothesis, not to manufacture exhaustive or repetitive copy.