Unverified50% confidenceFactExact time
The study broke down AI Overview answers into 98,020 atomic claims, of which 89% were explicitly or roughly supported, 7% were not mentioned in the cited text, and 4% directly contradicted the cited sources.
1
Sources
50%
Confidence
Long-term
Relevance
10/15/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedIn scenarios where AI-generated works account for a low proportion (e.g., only 5%), even with a detection accuracy of 90%, more than two-thirds of the flagged works are actually human-created—this is the base rate fallacy in Bayesian reasoning.71% similarUnverified研究表明在要求AI直接生成参考文献的任务中，幻觉率可高达40%-70%68% similarUnverifiedThe research reports output by AI range from 4,000 to 8,000 words in length, with 90% consisting of objective information.67% similarUnverifiedThe evidence counter shows: 5 inputs, 4 valid, 1 rejected, at most 4 AI calls.67% similarUnverified在案例测试中，直接将所有原始资料塞给AI的校验分为2/100，而经过上下文编译处理后的校验分为90/10065% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1035265API
curl https://kongchang.com/api/v1/knowledge/claims/1035265MCP
get_claim(id=1035265)