ResearchBenchmarksURL Health and Self-Correction on ExpertQAFor AI assistantsMarkdownContent indexURL Health and Self-Correction on ExpertQAFollow7,985Total URLsClaude4,051.725,072.866,0947,115.14Apr 3, 2026Evaluation resultsMethodMethodLinksTotal URLsAvg URLs per QuestionLive URL RateDead URL RateLikely Hallucination RateUnknown URL Status RateAvg Live URLs per QuestionMean Correction RoundsMedian Correction RoundsMax Correction RoundsClaudeAgentic tool usage=url...Agentic tool usage=urlhealth2026.047,98518.479.3100.420.214.64.0648GPT-5.1Agentic tool usage=url...Agentic tool usage=urlhealth2026.044,82911.178601.819.78.72.5226GeminiAgentic tool usage=url...Agentic tool usage=urlhealth2026.044,2039.788.9200.510.38.61.5622