September 9, 2025, at its fall event unveiling new iPhone, AirPods and Apple Watch models, Apple quietly introduced two major AI capabilities. ZDNET's hands‑on piece provides a direct appraisal: this is not merely a talking assistant, but generative AI deeply embedded into system workflows, changing how everyday information is processed.
The first feature is a system‑level text and image generation/understanding capability. The author used it on a long group chat thread with more than 20 messages; the system produced three key points, identified main participants and suggested three reply tones within five seconds. The interface appears as a floating AI entry point that opens to display various rewrite, summary and extension suggestions; when a travel photo is dragged into the interface, it can recognize landmarks in the image, offer stylized caption suggestions and automatically generate recommended social tags. The experience: no more copying and pasting across multiple apps—summarization and content generation are embedded in the original app, significantly improving efficiency.
The second feature focuses on voice: real‑time transcription, low‑latency translation and rewriting. In testing, the author recorded phone calls and demonstrated live translation; the system organizes spoken language into editable, sentence‑by‑sentence transcripts, supports converting a segment into a more formal or more casual tone, and can replay using localized voice synthesis. Importantly, most processing happens on‑device, with cloud models invoked only when the user explicitly agrees to enable more complex generation tasks; Apple's privacy statements are reflected in the hands‑on experience—latency is low without full reliance on the cloud.
The combination of both features creates notable scenarios: recordings can generate meeting highlights immediately, extract action items and automatically create tasks in Notes; on a business trip, the camera can scan a contract page and quickly draft a concise summary email ready to send. ZDNET notes the overall workflow treats AI more like an on‑call editor and translator embedded in apps, rather than a separate standalone app.
There are also limits in the experience: for highly specialized or deeply creative writing, auto‑generated suggestions still require human polishing; real‑time translation can still misinterpret dialects or strong accents. In addition, device model and chip differences directly affect the experience—the author notes some of the smoothest features require the next‑generation hardware.
ZDNET concludes the value of these two features lies not in isolated wow moments but in seamlessly integrating AI into daily operations, making writing, information organization and cross‑language communication readily accessible. For users who prefer efficient, immediate feedback—especially younger users—this change genuinely turns the phone into a smarter assistant that better understands their work.
It is apparent Apple is proceeding cautiously with AI integration: based on current steps and related reports, the guiding principle is to gradually incorporate features that are relatively mature and have good usability, while keeping less mature AI capabilities in testing until they reach sufficient maturity, such as the widely anticipated major upgrade to Siri.