AI search agents don't fail at searching, they fail at asking the right questions when queries get ambiguous

A new benchmark called DiscoBench tests AI search agents' ability to handle ambiguous queries by asking for clarification. Leading models like Gemini 3.1 Pro and Claude Opus 4.7 scored below 50 percent, indicating a need for improvement in handling uncertainty. The benchmark contains 211 tasks across eleven knowledge domains.

1sec.ai