How to build an AI-powered content-gap-analysis workflow

Search Engine Land how-to by Sara Vicioso (Director of Marketing, 12+ yrs B2B/B2C; edited by Angel Niñofranco, reviewed by Danny Goodwin) laying out a six-step content-gap-analysis workflow that hands the data-organizing work to an LLM (Claude) while keeping strategic judgment with the analyst.

The six steps

1. Choose competitors — 3–5 direct business competitors via Semrush’s Organic Competitors report; drop the noise (Amazon, Reddit, Wikipedia) and add stakeholder-named rivals Semrush misses.
2. Gather data — three sources: Semrush Keyword Gap (where competitors rank and you don’t), Google Search Console (validate existing authority + near-ranking queries), and Google Analytics 4 (engagement, conversions, sessions for business context). Fed in as CSV exports or via MCP (Model Context Protocol) direct connections.
3. Ask Claude to analyze — prompt it to “think like an SEO strategist” and cluster by intent, funnel stage, business relevance, and authority signals — not “cluster these keywords.”
4. Score and prioritize — a five-factor score (business relevance · existing authority · search demand · ranking difficulty · estimated effort) → High/Medium/Low, bucketed into quick wins, new content, or long-term authority plays.
5. Page-level recommendations — for high-priority items, Claude produces briefs (why selected, current rankings, recommended updates, expected impact, effort, priority) rather than keyword lists.
6. Measure — GSC (impressions, position, CTR) + GA4 (sessions, engagement, conversions), re-run quarterly.

The ~500-word prompt template takes business context up front and, notably, instructs the model to check its own work: “If reviewing Search Console or Analytics data changes your original recommendation, explain why.”

Why it’s here

A concrete instance of the seo-operating-model-shift at the task level: AI absorbs the commoditized row-scrolling and the human keeps the strategic call (content-gap-analysis). It also sits on the spoke’s clarity-over-volume side — the piece explicitly prefers a qualified-visitor topic over a high-volume-but-hard keyword. The stance that the analyst validates the model’s output against first-party GSC data echoes the spoke’s recurring measurement-from-the-source theme.

Cross-spoke context

The mechanism is an LLM (Claude) plus MCP tool-connections — agent-tooling machinery that lives in agentic-tooling-wiki (runner-up spoke). Routed here because the dominant substance is the SEO practice, not the harness; the MCP/Claude layer is the how, not the subject.

Tier

T3 — practitioner how-to on a trade publication (Search Engine Land), vendor-specific (Semrush + Claude), no primary data; useful as a documented workflow, not evidence. freshness: volatile — tool UIs and MCP connectors move fast.

Related

content-gap-analysis · seo-operating-model-shift · authority-density · google-search-console · search-marketing