Google's New Benchmark Reveals Wide Gaps in AI Factual Accuracy — and Shows Search Tools Help The best large language models get facts wrong about one-third of the time
