Content gaps
Find prompts you're losing and the URLs the model trusts there.
The Content gaps view is the recommendation engine of the product. Citations tell you what’s happening; gaps tell you what to do about it.
What counts as a gap
A monitored prompt is a “gap” when:
- Your brand isn’t mentioned in the answer.
- The model cited at least one URL — i.e., there’s competitive content to learn from.
- The prompt has been running long enough to trust the result (typically three runs with consistent behaviour).
Gaps are scored by how often the prompt is asked and how visible the opportunity is (citation density, competitor concentration). The highest-scoring gaps are the ones to fix first.
Reading a gap
Each gap row expands to show:
- The prompt and the personas / tags attached to it.
- The current answer (a representative recent run).
- The URLs the model cites and the brand behind each.
- Suggested action — typically one of: write a page that targets this prompt, pitch a guest post on one of the citing publications, or update an existing page to match the angle the model rewards.
From gap to brief
For accounts on Pro and above, you can generate a content brief from any gap. The brief includes:
- The prompt and its variants.
- A factual outline based on what the cited URLs actually cover.
- Suggested headings, key facts, and entity coverage.
- Voice and tone guidance pulled from your brand-book.
Briefs are markdown; copy them into Notion, Google Docs, or your CMS. They’re a starting point, not a finished article — keep your editorial voice.
Closing the loop
Once you publish, attach the new URL to the gap (the row has a “Mark addressed” action with a URL field). The next run on that prompt will tell you whether the page got cited. If it did, the gap closes. If it didn’t, you’ll see why — maybe the angle was off, maybe the publication hasn’t been re-crawled yet by the model, maybe a competitor still owns the citation slot.
API
-
GET
/v1/projects/{projectId}/content-gaps -
GET
/v1/projects/{projectId}/content-gaps/{gapId}
Cursor-paginated, sorted by score descending by default.