Apple’s rebuilt Siri AI is meant to feel less like a command box and more like a useful layer across the iPhone. A hands-on test around San Francisco suggests that change is real, but still uneven in beta.
The assistant handled travel questions, searched personal photos, opened the camera, drafted messages, and used local context to make recommendations. It also made mistakes that show why Apple still has work to do before the public rollout later this year.
A More Present Siri On The iPhone
The new Siri AI arrives as part of iOS 27 and represents a clear shift from the older Siri experience. Instead of sitting apart from the rest of the phone, it is merged into the iPhone search bar and appears when the user swipes down in the middle of the screen.
The interaction model is also broader. Users can speak conversationally, then continue by swiping down on Siri’s answer and typing follow-up questions. Those exchanges are saved in a dedicated app, making it possible to revisit earlier conversations rather than starting over every time.
In the San Francisco test, Siri’s responses were usually short and scannable. When asked for a beach hike route near the Golden Gate Bridge to see the sunrise, it recommended a popular trail in the Presidio neighborhood and another option in the Marin Headlands. Its answer appeared as both speech and text, with important words bolded for quick reading.
That matters because travel help often fails when it gives too much information at once. Here, Siri AI appeared designed for fast decisions: ask, scan, follow up, and move on.
Personal Context Is The Main Upgrade
The biggest change is personalization. Apple says the new Siri can draw on messages, photos, and emails to answer questions in a more relevant way. In practice, that means the assistant is not only searching the web or offering links; it can look across what is already on the device.
When asked a broad question such as what to do that day, Siri reviewed recent messages and surfaced plans that had been discussed with friends but not finalized. That is a different kind of assistant behavior: less generic suggestion, more personal reminder.
Apple’s partnership with Google is described as a core part of this overhaul. Google’s Gemini helps power the underlying model, Apple Intelligence. In the hands-on test, that model made Siri’s output feel more connected to the user’s actual intent rather than simply pointing to websites.
The assistant also showed flexibility across services. When asked to draft a text, Siri asked whether it should send the message through Apple’s Messages or Meta’s Messenger service. That detail suggests Apple is positioning Siri AI as a phone-wide assistant rather than only an Apple-app feature.
Indexing And Privacy Shape The Experience
For Siri AI to answer personal questions, the iPhone needs to index the data on the device. That means scanning and cataloging information so the assistant can retrieve it when asked.
In the developer beta test, the iPhone took a little over a week to fully index after being updated to iOS 27. That delay is important for users to understand: the personalized version of Siri may not feel complete immediately after installation.
Apple emphasized privacy at WWDC 2026. The company referenced its Private Cloud Compute approach and claims it does not store user data, only pulling from it when someone asks Siri a question. Users who do not want the feature can turn off Siri AI in settings, similar to the previous version of Apple’s assistant.
Hardware support will also vary. The test used an iPhone 16 Pro Max, which will get many but not all Siri AI features. Based on what has been publicly released, only the iPhone Air, iPhone 17 Pro, and the iPhone 17 Max will have all the fixings, including more varied voice options. Every iPhone 16 and iPhone 17 model will be able to run the new Siri, while only the iPhone 15 Pro and Pro Max will be compatible. Older models will not support this voice assistant.
Where Siri AI Helped During The Day
As a travel guide, Siri AI handled several practical tasks well. Near the Golden Gate Bridge, it responded to hiking questions with brief route ideas. Later, when asked for fluffy pancakes nearby, it suggested brunch options with online reviews and map directions.
Siri highlighted one restaurant as a standard brunch choice with cozy vibes and another as a more retro-inspired experience. The tester chose the cozy option and found the recommendation decent, even if the price was high.
The assistant also worked inside the camera flow. When asked to take a selfie, Siri opened the camera app, counted down, and captured an image with the front camera. Afterward, it was able to start drafting a message to Sam with the photo attached.
Photo search was another strong area. When asked to find photos from the last time the tester went to Costa Rica, Siri found images from two years ago and displayed them inside the Siri app. For anyone with thousands of photos, that kind of retrieval could be one of the most useful everyday upgrades.
The Beta Still Shows Rough Edges
The same test also showed why Siri AI is not finished. When the camera was pointed at a foggy path, Siri identified Monterey cypress trees but responded with a short history of the Cypress Tree Tunnel at Point Reyes National Seashore. That location was an hour drive away from where the tester actually stood, which could confuse someone unfamiliar with the area.
Photo search also produced mixed results. A request for photos of hot pot with friends surfaced images of wagyu beef, which matched the intent. But it also returned vacation images from a hot tub, showing that the assistant can still misread similar language.
Messaging had its own problems. Siri found Sam’s contact information and started drafts correctly, but it treated dictation too literally. It included the words “with a” before the skull emoji, once used a school emoji instead, and once asked whether the message should also go to Adam.
These are not small issues for an assistant that will be trusted with personal communication and travel guidance. The promise is clear: a Siri that understands context, takes action across apps, and helps without long explanations. The beta reality is that Apple still needs to tighten accuracy, intent recognition, and edge cases before Siri AI can feel dependable in public use.