Analyzing Content Gaps

Overview

The top 3 pages already won the editorial battle for a query — Google watched users vote with dwell, scroll, and pogo-sticks for months and decided those pages best serve the intent. Your job isn't to imagine what the page should cover. It's to diff what they cover that you don't, then decide which gaps deserve a response.

Core principle: SERP-proven coverage beats imagination. Entities and sub-questions present across all top-3 are table stakes — if they're missing on your page, fix those first.

Why this matters in 2026

• AI Overviews appear on 50–60% of US Google searches, average 157 words, and cite 8–13 sources [3]. AIO citations skew toward pages with deep entity coverage on the specific sub-question — 92% of citations come from top-10 pages, but 2 in 3 come from pages a normal SERP search wouldn't surface [4].
• Topic Gaps replaced keyword gaps as the unit of analysis [1]. Modern search treats keywords as surface representations of underlying entities.
• Featured snippets and AIO citations correlate strongly — the answer-first patterns that win snippets win citations [3].
• Skyscraper 1.0 is dead. "Make it longer" no longer works; the 2026 evolution is better entities, sharper examples, tighter intent — not word count [5].

When to use

• A competitor consistently outranks you for a query you care about
• You're refreshing a page and need to know what to add
• You need a content brief before writing new
• A page sits at position 5–15 and finding-underserved-keywords shows the gap is structural, not striking-distance
• You want to know why your competitor is in the AIO and you aren't

Don't use for: brand-new sites with no SERP presence (use researching-keywords-pre-launch), pages whose problem is decay (use refreshing-stale-content), or queries where SERP intent doesn't match your page type — fix intent first.

What to compare — the diff axes

The analysis flow

1. Inputs. Target query + your URL. Confirm the query has a SERP and your intent matches the live top-3.
2. Scrape top-3. Use firecrawl MCP — firecrawl_search for the SERP, then firecrawl_scrape per result with formats: ["markdown", "links"] and onlyMainContent: true. Skip ads/sponsored/aggregators unless they are the intent. Firecrawl respects robots.txt by default [2].
3. Extract entities per page. Two options:

• Higher accuracy: Google Cloud NLP analyzeEntities (~$1/1k requests) — named entities with salience scores [7].
• Free / good enough: noun-phrase frequency via spaCy or compromise.js, deduped by lowercased lemma. Gets ~85% of the signal [7].

1. Extract structure — H2/H3 tree, schema types, word count, dateModified, internal-link count.
2. Build the diff matrix. For each axis, list each competitor vs. you. Highlight: entities all 3 have but you don't, H2s 2+ share, schema overlap, depth/freshness deltas.
3. Rank gaps:

• Tier 1: Entity or H2 in all 3 competitors, absent on your page.
• Tier 2: Schema gaps with rich-result or AIO upside.
• Tier 3: Depth/freshness — your page <80% of median word count, or dateModified >9 months stale.
• Tier 4: Single-competitor entities — usually noise, unless that competitor is the AIO-cited one.

1. Output `CONTENT_BRIEF.md` with: query, your URL, top-3 URLs, diff matrix, ranked gaps, expand-vs-write-new call.

Expand vs. write new — decision tree

• Gaps are entity-level within existing scope → expand the page. Hand off to optimizing-on-page.
• Gaps span 2 or fewer sub-topics that fit the page's intent → expand with new H2 sections.
• Gaps span 3+ sub-topics or shift intent (informational ↔ transactional) → write new spokes. Hand off to planning-topic-clusters.
• Gaps are schema-only → hand off to adding-schema-markup.

Concrete example: vibecrafting.ai for "capsule wardrobe formula"

Top-3 are NYT, Vogue, a Substack. Diff reveals:

• All 3 cover entities seasonal palette, core 4 silhouettes, 70/20/10 ratio — your page mentions zero. Tier 1.
• 2 of 3 have an H2 "How many pieces should a capsule wardrobe have?" — yours doesn't. Tier 1.
• All 3 have FAQPage schema; you have only Article. Tier 2.
• Median word count 2,400; yours is 1,100. Depth gap, but only act after Tier 1 — adding words without adding entities is the skyscraper trap [5].
• NYT is cited in the AIO with a 45-word answer under a question-phrased H2; your page buries the answer in paragraph 4.

Recommendation: expand (one page worth of edits, not a new spoke). Hand off to optimizing-on-page with this brief.

Common mistakes

• Copying competitors verbatim — Google flags near-duplicate phrasing; AI systems deprioritize derivative content. Cover the same entities, write your own sentences.
• Ignoring intent — a transactional competitor in a mostly-informational SERP is noise.
• Gap-stuffing — covering everything makes the page worse. Stop at Tier 1 + Tier 2.
• Forgetting your differentiator — your unique angle is the spine; gap entities weave around it.
• Diffing text only — schema and freshness gaps are invisible to a markdown diff but often the highest-leverage fix.
• Running this on a page with no SERP yet — use researching-keywords-pre-launch instead.

Quick reference — gap type → response

Lifecycle awareness

Growth-phase skill (3–12 mo). Skip pre-launch (no SERP signal yet — use researching-keywords-pre-launch). For mature pages, pair with refreshing-stale-content — the gap may be freshness, not coverage.

What next

• Tier 1 gaps to fill → optimizing-on-page with the brief in hand.
• Multi-subtopic gaps → planning-topic-clusters.
• Schema-only → adding-schema-markup.
• AIO citation specifically → optimizing-for-generative-engines after the page is structurally sound.

Citations and verification tags in SOURCES.md.