Context
Sigillo and colleagues introduce EMBL AI Librarian, a life-sciences knowledge layer that plans complementary subqueries over live Europe PMC search, reads selected papers and locates evidence for grounded answers. Their original preprint reports improvements on citation and question-answering benchmarks.
What changed
The authors released a domain-specific agentic retrieval approach combining subquery planning, live scholarly search, evidence localization and benchmarked citation-quality outcomes, with an accompanying code repository.
Why it matters
This provides direct research evidence for separating multi-query retrieval, source selection and passage-level evidence grounding when analyzing AI discovery pipelines; it does not measure public-web publisher visibility.
Visibility OS interpretation
Visibility OS interpretation: AI discoverability measurement should distinguish whether an agent retrieves a source from whether it locates supporting evidence and cites it correctly. This interpretation is not an observed ranking rule for ChatGPT Search, Google AI Mode, Gemini, Perplexity or Bing.
What remains uncertain
How well do the reported gains replicate independently and outside the life-sciences corpus? Would similar methods change citation exposure in live commercial web-search products? How do source selection and citation accuracy vary under different retrieval backends?