Content gaps

What is a content gap?

A content gap is a topic where AI answers currently favor a competitor over you — either because you have no page that answers the question at all, or because the page you do have isn’t convincing enough to get cited. Gaps live on the Generate screen, ranked by a deterministic evidence index that helps you decide what to inspect first.

How gaps are found

Gaps come from three independent detectors that all read the same source: trusted answers captured for your tracked prompts. A signal must appear in the newest two distinct tracking batches for the relevant prompt and platform; recomputing the same batch cannot confirm it twice. Each detector looks for a different kind of absence, and overlapping evidence is merged into one card per topic.

1. Losing prompts

A tracked prompt counts as “losing” on a platform when, in each of its two latest independent batch observations, your brand is neither mentioned nor cited and at least one competitor is. If your brand shows up at all — even just a mention with no citation — that observation is not a loss. Losing prompts are grouped (“clustered”) under the project topic their wording best matches; prompts that don’t match any of your defined topics are grouped together under a label built from their two most common shared words.

2. Competitor citation gaps

This detector ignores whether a prompt is classified as losing, but it still uses citations collected from tracked-prompt answers. Every competitor URL cited across both independent batches is grouped by its normalized address. A group becomes a gap only if none of your existing content pages match it closely enough using the page’s title, headings, lead, meta description, and body text, with guards against a single incidental shared word — in plain terms, they have a page for this and you don’t. Pages that already cover the subject are filtered out here.

3. Cluster completeness

For each topic in your brand profile, this detector looks at how many distinct subtopics competitor citations cover under that topic. If competitors have citation evidence across two or more subtopics and you have no page that matches the topic overall, it’s flagged as a coverage gap — you’re not just missing one page, you’re thin on the whole subject. The card lists up to six of the uncovered subtopics by name.

The Opportunity Score

Every gap gets a single number from 0 to 100 — the Opportunity Score — so you can work the queue in a useful order without reading every card in full. It is an in-product prioritization index, not a probability that content will rank, earn a citation, or produce a particular lift.

The four signals it blends, in the order they carry the most weight:

Why prompt category matters

The category term is a fixed product prior, not a measured success rate. It gives more weight to category and comparison questions than to prompts already centered on one brand:

1. Unbranded — top-of-funnel questions with no brand named yet; assigned the highest prior.
2. Comparison — high-intent and still up for grabs.
3. Informational — useful, broader-interest questions.
4. Branded — already about a specific brand; assigned the lowest prior.

Unbranded and comparison prompts score highest because they’re treated as higher-intent, less brand-locked territory. A gap takes the highest-ranked category among its losing prompts, not an average, so one unbranded prompt is enough to lift the whole cluster.

Reading a gap card

Every gap card leads with its evidence, in the same order it was detected, before any call to action:

• Kind badge — Losing prompts, Competitors cited, or Topic coverage — plus the topic name and the Opportunity Score in the corner (colored green at 70+, amber at 40–69, slate below that).
• Losing prompts — the exact prompt text, which platform(s) it’s losing on, and which competitors show up instead (mentioned or cited — hovering a competitor chip tells you which).
• Cited instead — the specific competitor URLs earning citations on this topic, each linking straight to the page, with how many times it’s been cited and which competitor it belongs to.
• Uncovered subtopics — for coverage gaps, the subtopic labels competitors have covered that you haven’t.

Only after that evidence do you see the action row. Open rows are split into Ready to act, Needs validation, and Not recommended. Create brief is available only after validation says the gap is actionable. Pending rows remain visible with the action disabled; not-actionable rows remain visible, are capped at 25, and are excluded from suggested work. Every open row can still be Dismissed with a reason (Not relevant, Already covered, Low value, or Other).

The gap lifecycle

Gaps are never deleted — they move through four statuses, and the Generate page shows a tab with a live count for each:

From gap to brief

Once you’ve reviewed the evidence and validation marks the row actionable, Create brief turns the gap into a fully editable content brief — the working title, outline, target queries, and everything else you’d need to write the piece yourself, generated automatically from the gap’s evidence. Creating a brief doesn’t commit you to anything irreversible: the brief starts as a draft you can rewrite before any article gets generated. See Briefs for what it contains and how to work with it.