Triage content gaps
The Content Gaps tab is the single home for "content readers want but cannot find." It pulls together two signals from your published docs and lets you act on each one directly.
Before you begin
You need a published site with reader traffic on its AI search or chat widget.
Open the tab
Go to Analytics in the sidebar.
Open the Content Gaps tab.
The amber Zero-result rate card on the AI queries and crawler activity tab also links here.
Suggested gaps
The Suggested gaps section lists clusters that weekly gap detection found. It groups reader AI queries that returned zero or poor results, clusters similar wordings together, discards off-topic noise, and proposes a topic title for the rest. Each row shows:
The suggested topic title (or the representative query, if no title was generated).
How many queries are in the cluster and the representative query text.
A confidence badge (Strong, Likely, or Possible) reflecting how often the queries occurred, how recent they are, and how unlike any existing topic they are.
For each suggestion you can:
Dismiss (the X): remove a suggestion that is off-topic or already covered.
Create topic (the +): create a draft topic seeded with the suggested title and open it in the editor. The gap is marked as addressed.
Triage strongest-confidence gaps first. A Strong gap means several recent readers asked something your docs do not cover at all, the clearest signal to write a new topic.
Searches with no answer
The Searches with no answer section lists raw reader searches that returned nothing: the unprocessed demand signal, before clustering. Each row shows the query, how many times it was searched, and when it was last asked. Click Create topic to start a draft with the search as its working title.
Not every gap needs a new topic. Some queries are off-topic or outside your documentation's scope, and some are better answered by improving an existing topic than by adding a new one. Use the suggested title and query count as guidance, not a mandate.
Where gaps come from
Gap detection runs automatically each week for eligible projects, analyzing reader AI queries from the last 30 days. It only considers queries that returned no usable answer or drew negative reader feedback, and it needs a cluster to recur before suggesting it, so single one-off questions do not create noise.
See also
Prioritize what to write next: gaps and content-health signals ranked into one worklist
AI queries and crawler activity: what readers ask the on-site AI
Monitor search analytics: keyword search gaps from the search widget
Improve AI search readiness (GEO): make existing pages easier for AI engines to cite