How to Run a Google Search Console Content Gap Analysis (Step by Step)

A content gap analysis answers one question: which queries is Google already showing your site for, that you have no page actually targeting? Those are the cheapest wins in agentic SEO, because the site already has some relevance signal for the query — it just hasn't been given a dedicated page to earn the click. This is the export-driven version of the process, usable with nothing more than a Search Console export and a spreadsheet, whether or not you're running it through an AI agent.

Quick answer: run a content gap analysis by exporting Queries + Pages data from Search Console (last 3 months), filtering for queries with real impressions but no matching dedicated page, then cross-checking for cannibalization (two pages competing for one query) before creating anything new. Prioritize by impressions × buyer relevance, divided by estimated difficulty.

SEO CEO runs this exact workflow automatically inside your agent, if you'd rather not do it by hand every week.

What data do you actually need?

One export: Search Console → Performance → Search results, filtered to the last 3 months, with both the Queries and Pages tabs exported. You want impressions, clicks, average position, and CTR for each query, plus the same for each page. Three months balances recency against enough volume to be statistically meaningful — a shorter window is noisy, a longer one hides recent shifts.

How do you find a genuine keyword gap?

Sort the Queries export by impressions, descending. For each query with meaningful impressions, check whether any live page is actually built to answer it — not just ranking for it by accident, but structured around it with a matching H1, headings, and intent. If nothing qualifies, that's a gap. The signal that makes this worth doing over guessing keywords cold: Google is already showing your site for the term, which means there's some existing relevance to build on rather than starting from zero.

How do you catch cannibalization instead of creating a duplicate?

Before building a new page for a gap, check the Pages export for that same query. If two or more pages are already getting impressions for very similar terms, that's cannibalization, not a gap — and the fix is to merge, differentiate, or canonicalize the weaker page, not add a third competing page. Skipping this check is the most common way a gap analysis makes a site's cannibalization problem worse instead of better.

How do you prioritize which gaps to act on first?

Not every gap deserves an article this week. A simple, workable formula: priority = (impressions × buyer relevance) ÷ difficulty, where buyer relevance is a rough multiplier — 0.5 for general informational interest, 1 for topically relevant, 2 for a query someone types when they're ready to pay. Sort by that score and take the top 5–10 for this cycle rather than trying to clear the whole backlog at once.

What do you do with the prioritized list?

Turn the top opportunities into either a fix (if a related page already exists and just needs to be sharpened) or a new article brief (if nothing exists yet), using an answer-first structure with a direct answer near the top of the page. Ship the highest-priority 3–5, submit the new or updated URLs for indexing, and log what shipped so it feeds next week's export.

How often should you repeat this?

Weekly is the cadence that compounds: each page that starts ranking brings impressions for adjacent queries, which shows up as next week's gaps. Doing this monthly instead of weekly doesn't make each session easier — it just means the compounding effect is slower to show up.

What does a finished gap-analysis output actually look like?

In practice it's a simple table, not a dashboard: one row per opportunity, with columns for the query, current impressions, average position, the page (if any) currently getting the impressions, whether it's a genuine gap or cannibalization, the priority score, and the action — fix, differentiate, or create. Keeping it in a flat table rather than scattered notes is what makes the weekly version of this fast; most of the second and later cycles are updates to the same table, not a rebuild from scratch.

If you're running this through an agent rather than by hand, this maps directly onto the weekly gap-analysis loop described on the how it works page, and pairs naturally with an AEO/GEO pass on whatever ships.