Skip to main content
GET
Each row is a question that is a gap in your content, for one of two reasons. abstained: the assistant said it could not answer from its sources rather than guess. rated_unhelpful: the assistant answered, and the person who asked rated the answer down, in the console or in the website widget. See How answers work. Rows carry the question text but never the answer, the citations, or who asked. total counts every row in the window for both reasons, even when limit returns fewer. Send group=true to get the same rows arranged into groups of similar questions, in groups instead of questions. Two questions share a group when they mean the same thing, or when at least four fifths of their words match after case, punctuation and common English words such as “how” and “the” are set aside. grouping says which comparison was used: when it is wording, two questions that mean the same thing in different words are in separate groups. Grouping covers only the rows this request returns, so a group’s count is a count within the page, not within the window. Every read is recorded in your organization’s audit trail, grouped or not. Because this returns other people’s questions, it requires the owner or admin role. Editors, members and dataset operators receive 403, even though editors can read usage counts.
string
default:"30d"
Trailing window ending now: 7d, 30d or 90d. Any other value returns 400.
string
Limit the results to one assistant (UUID). A malformed ID returns 400; an ID from another organization matches nothing.
integer
default:"50"
Maximum rows to return, from 1 to 200. ASCII digits only.
string
default:"false"
true returns groups of similar questions in place of questions. false returns the plain list. Any other value returns 400.

Response

object
required
The window searched.
integer
required
How many rows match, for both reasons, regardless of limit. A row whose question is no longer stored is counted here but not listed.
object
required
total split by reason.
object[]
Unanswered questions, newest first. Present unless group=true.
string
Present only with group=true. How the groups were formed. meaning: every question was compared by meaning. partial: some were compared by wording only. wording: all were compared by wording only. Treat a value you do not recognise as partial.
object[]
Present only with group=true. Every returned row is in exactly one group, and a question unlike any other is a group of one. Largest group first, then the most recently asked.
Last modified on November 7, 2026