Apple Watch Audio Intelligence Tests Reveal Siri Recap Flaws
Testing the new Audio Intelligence features on the Apple Watch Series 12 and Ultra 4 reveals mixed results with Siri Recap and Live Rewind.

Recent hands-on testing of the Apple Watch Series 12 and Apple Watch Ultra 4 provides a closer look at Audio Intelligence, a newly introduced set of features designed to process daily conversations. Operating within the WatchOS 27.3 and iOS public beta environment, the capability includes tools like Siri Recap, which generates daily summaries, and Live Rewind, offering a transcript of the preceding 15 seconds of audio.
Unlike dedicated AI-powered recording pins, these capabilities come built directly into the smartwatch hardware at no additional cost. Users must explicitly opt-in to activate the system, retaining the ability to run it on a schedule or toggle it manually throughout the day. Reminders help users keep track of when the software is actively listening.
Siri Recap Performance and Privacy Concerns
While the option for manual control helps minimize the intrusive feel of continuous monitoring, observers note potential privacy considerations regarding individuals within earshot. Unlike smart glasses that feature exterior recording indicator lights, the Apple Watch relies on a subtle, barely audible chime when a user triggers Live Rewind via the Digital Crown, leaving nearby spectators largely unaware.
In practical application, the utility of Siri Recap varies. During focused work sessions—such as following a designated lab stream where Matthew Murray, CNET’s lab director, discussed folding phone durability metrics—the feature successfully captured specific data points. However, testing also exposed inaccuracies, including mistaken names and details during car-pool logistics conversations, as well as instances where Siri conflated unrelated discussions or misinterpreted television dialogue from a Netflix show as an active personal conversation.
Ultimately, while these software additions offer an interesting glimpse into wearable AI development, early evaluations suggest they remain supplementary rather than essential features for everyday device operation.
