Answer Generation
Where it writes the answer, does it say only what its own sources support -- declining when they don't, flagging when they disagree, and attributing what it does claim? Judged against what THAT TOOL retrieved, never against the web. Out of scope for search-only products, which is not a zero and never drags them down a ranking. A citation that correctly points at the source of a wrong answer is a CORRECT citation.
What this capability means
For answer-generating tools, this capability means the final answer stays within the tool’s own retrieved support. When the retrieved sources do not support an answer, the tool should decline; when sources conflict, it should flag the disagreement and attribute claims to sources. The answer is judged against the tool’s retrieved sources, not against the live web.
Boundary: This does not judge search-only tools that return results without writing an answer.
Scenarios that test this capability
A scenario is a real-world situation used to test a capability.
Benchmarks that include this capability
A capability has global identity and may be used by more than one benchmark.
Web Search for AI
Which web search or answer API finds the live-web material that actually supports an answer, keeps up when that material changes, and -- where it writes the answer itself -- says only what its own sources support?