Finding and Closing Content Gaps

A content gap is a question your buyers ask where an AI engine answers without you. The Content Gaps page finds them from two directions, and the two halves are complementary rather than alternatives: one tells you which questions you are missing, the other tells you which pages are too weak to win them.

The page lives in the Optimise group of the sidebar and opens on two tabs, Query gaps and Site audit.

Step 1: Read the query gaps from your latest scan

The Query gaps tab derives everything from your project’s most recent completed scan. For every prompt that scan ran, it records which engines answered without mentioning your brand, and groups the misses by question.

A gap is a content problem before it is a visibility problem, and the fix is usually the ordinary one: Google’s creating helpful, reliable, people-first content guidance describes the same qualities that make a passage quotable by an answer engine: a direct answer near the top, self-contained sections, and evident first-hand expertise.

Four stat cards sit on top:

• Content Gaps Found: queries with at least one miss.
• Missed Mentions: query × engine misses, so a question missed by four engines counts four times. This is the number that shows severity; the first shows breadth.
• Engines With Gaps: how many distinct engines are missing you somewhere.
• Gap Trend: how the gap count moved against the previous scan. A negative number is the product working: queries left the table since the last scan.

Below them, the Gap Analysis table lists each query, ordered by opportunity, with the context that tells you which gap to work first:

• Volume: the query’s estimated monthly searches and its priority tier, taken from the matching tracked prompt (see the Prompts page). A gap you have never heard of stays a list entry; a high-volume gap is a decision.
• Engines Missing You: badges for the engines that answered without mentioning your brand.
• Competitors Instead: who the engines named in your place. This is the AI analogue of the classic keyword gap, computed from the same answers you already paid for.
• Google: when Search Console is connected, whether Google ranks the query on page one even though no engine mentioned you, with the impressions proving the demand. A query ranking in Google while every engine ignores you is the strongest version of this whole page.
• Opportunity: a transparent 0–100 score: the query’s demand (its volume or Google impressions) times how many engines missed it. No score appears when there is no demand signal at all. A gap with no data is reported, never invented.
• Fix: jumps to the GEO recommendations page, which turns these very gaps into suggested actions.

Read the engine badges, not just the query. A question missed by one engine is a model quirk. A question missed by every engine is a content problem you can fix.

If your brand was mentioned everywhere, you get a clean empty state saying so: a real result rather than a failure to load.

Step 2: Tell a real gap from a sample preview

Before your first completed scan, this tab shows illustrative sample data (a filled table with plausible-looking queries, competitors and opportunity scores) behind a notice that reads “Sample data shown for preview. Run a scan to see live content gap analysis for your brand.”

Take that notice seriously. The sample rows exist so the layout is not an empty rectangle on day one; they are not about your brand and never touch your data. Two tells give the preview away: the sample table carries a difficulty column, which real results never measure, and its competitor names and opportunity scores are invented. Real gap analysis draws its volume, priority and opportunity from your tracked prompts and Search Console, so before either exists you will see plain queries with the engines missing you.

The fix is simply to run a scan. Everything on this tab reflects the latest completed one, so a gap you closed last week disappears from the table only after the next scan confirms it.

Step 3: Turn a gap into a hypothesis

A gap row tells you what happened, not why. Before writing anything, decide which of the usual causes you are looking at:

• No page exists for that question at all. The commonest case and the easiest to act on.
• A page exists but does not answer the question directly: it circles the topic without ever stating the answer in a form an engine can lift.
• The page exists and answers well, but nothing establishes that you are a credible source for it: no citations, no entity signals, no corroboration elsewhere.

Each cause has a different fix, and the second tab is how you tell the second and third apart from the first.

Step 4: Run a site-wide GEO audit

Switch to Site audit. Before a first run you get an empty state and a Run site audit button; afterwards the button reads Re-run audit, and Auditing site… while one is in flight.

What the run does, in order:

1. Discovers your pages. It looks for sitemap.xml first and falls back to a shallow crawl of the site itself. Both paths are bounded and both honour robots.txt.
2. Selects the pages worth grading: up to ten per run. Selection is not arbitrary: duplicate URL variants collapse, utility and legal pages are dropped, documentation, pricing and FAQ-style paths are boosted, and pages matching your high-volume prompt topics are boosted again. That last input only exists once you have run Discover prompts on the Prompts page. The topic clusters and their volumes are written by that run, not by the prompts you add by hand, and a project that has never run one gets a selection identical to a project with no prompt work at all. Where topics do exist, the boost is worth up to 40 points against a path-importance score that spans roughly 30 to 130, so it reorders pages that path importance had scored close together rather than deciding the ten outright.
3. Grades each page for how well an AI engine can understand, trust and cite it.
4. Aggregates the results into a site score and a fix list.

Three constraints to know. Site audits require a paid plan: every graded page is a real analyzer call. Viewers cannot start one. And one audit per project per hour is allowed: click Re-run audit inside that window and the run is declined with a red banner saying an audit already ran in the past hour, rather than quietly re-using the previous result. The window runs from the last audit’s start, so a run that finished twenty minutes ago still blocks the next one.

Step 5: Read the audit scores

Four stat cards summarise the run:

• Site GEO Score: priority-weighted, 0–100. Weighted, not averaged, so a weak home page hurts more than a weak leaf page.
• Pages Graded, shown against how many were discovered.
• Pages Unreadable: pages that failed to fetch or grade. They are counted honestly and never dilute the score, so an audit where half the site was unreachable cannot look like a good one.
• Discovery: whether the run used your sitemap or fell back to a crawl. A “Crawl” here is worth fixing on its own: if we could not find a sitemap, some AI crawlers will not either.

Category scores breaks the result into the dimensions the grader works in: structure, clarity, entity coverage, answer-readiness, and metadata/schema. This is the fastest read on the page: a site that scores well on clarity and badly on metadata has a very different backlog from one where the reverse is true.

Page scores lists every graded URL with its matched topic, its priority and its score, and marks any page that could not be graded with the reason.

Step 6: Work the fix list in priority order

Top fixes across your site collects the most common concrete improvements the grader proposed across the pages it read. Work it top-down rather than page-by-page: a structural fault that appears on eight pages is worth more than a perfect rewrite of one.

Then close the loop with the first tab:

1. Pick a gap query whose engines nearly all missed you.
2. Find the page that should have won it in the Page scores table, and read its score and topic.
3. Apply the relevant items from the fix list to that page.
4. Wait for the next scan, and check whether the query has left the Gap Analysis table.

That last step is the discipline that makes this page worth using. A gap is only closed when a scan says so, not when the page is published.

Step 7: Where to go next

• Content gap analysis guide: the method behind the table, including how to prioritise between gaps.
• Content optimization for AI: what “answer-ready” actually means at the page level.
• Writing for AI citation: how to state an answer so an engine can lift it.
• GEO recommendations: the feature that turns a scan straight into suggested actions.