Control panel → Questions — user questions that LiKE answered poorly or couldn't answer at all. Each such question is a hint about which article is missing. This section is available to the Editor role and above.


#What counts as a gap

A gap is a search query whose answer covered the topic by less than 75% (the coverage score is assigned by the LLM when the answer is generated). Such queries are collected automatically from the search history for a period and land in the "Questions" section.

Why this matters: to see real gaps in the knowledge base, prioritize content by actual demand, and understand what employees search for and in which words.


#Triaging questions

A list of questions with sorting (by frequency / by date / by AI confidence) and filters: status (all / resolved), section, clusters. Each question shows its text, how many times it was asked (×N), the language (flagged if the query is cross-language), the section, and the time of the last request.

Each question is resolved through one of two tracks:

✦ AI judgment — LiKE analyzes the question and either:

  • suggests an existing article with a confidence percentage and rationale — the "Accept" button;
  • recommends creating a new article if no suitable one exists.

You can run or re-run the analysis manually with the AI analysis button.

Your decision — manually:

  • Assign to an existing article;
  • Create a new one (opens the editor);
  • Archive the question.

Similar questions can be combined into clusters and closed with a single article for the whole cluster. Bulk selection of questions via checkboxes is supported.


#Where else gaps are visible

  • Overview — the "AI recommends" banner with unreviewed gaps.
  • Editor — the "Insights" panel: questions related to the current article (ranked by relevance to its title and content) and a coverage check. This closes the loop: question → creating or improving an article → closing the question.