Context
An original preprint audits Google AI Overviews for 3,611 benchmark health questions, comparing cited sources with traditional Google results, source credibility, statement grounding and physician-written accuracy rubrics.
What changed
Introduced a domain-specific joint assessment of source selection and claim grounding; authors report only 44.4% of top-five conventional results were cited by corresponding AI Overviews and estimate 56.7% of cited statements were supported by cited pages after classifier correction.
Why it matters
Shows why AI citations, source credibility and answer-grounding must be measured separately when assessing AI discoverability and reliability.
Visibility OS interpretation
Visibility OS interpretation: being selected as a source is not equivalent to being accurately represented or used as evidence. The study supports multi-stage measurement, not an actionable ranking-factor claim.
What remains uncertain
How robust are classifier-error corrections and physician rubrics? Do results generalize beyond the sampled medical queries, time and geography? The 96.4% AIO occurrence is benchmark-specific.