How the Gemini layer actually works
The two headline features share a design pattern: Gemini turns free-form speech into structured actions, but keeps a human or a database in the loop rather than acting autonomously. Conversational Reporting now parses a spoken cue like 'The road is closed here' and packages it into a map-update suggestion that is routed to local map editors, who verify it before the change goes live for everyone [1]. Destination search works the other way around, converting a spoken query such as 'Find me a gas station nearby with the lowest prices' into a lookup against Waze's existing data on hours, parking, and fuel prices [1]. In both cases Gemini is the natural-language front end, not the source of truth - the verification step for map edits is what preserves data quality even as the reporting surface gets easier to talk to [3].


