Which Siri 2.0 features have actually shipped, with Apple's own example phrasings - and which headline abilities are still only in the iOS 27 beta.
Quick answer: Siri today is genuinely better at follow-up questions, recovering from stumbles, answering questions about your Apple devices, and typed requests. The three headline features people actually want — onscreen awareness, personal context, and in-app actions — have not shipped yet. They are in the iOS 27 beta. Everything below is separated into what works right now and what does not, because that gap is the whole story.
Source check: August 31, 2026. Shipping examples are taken from Apple's own iPhone User Guide for iOS 26; unreleased examples are from Apple's June 2026 WWDC announcement of Siri AI in iOS 27, which Apple describes as a beta. Current shipping iOS at the time of writing is iOS 26.6.1; iOS 27 is in beta.
What "Siri 2.0" Actually Refers To
The confusion here is real and it is not your fault. Apple announced a rebuilt Siri at WWDC in June 2024 with three marquee abilities: understanding what is on your screen, drawing on your personal data, and taking actions inside apps. Those are the features that made the demos impressive.
They did not ship in iOS 18. They did not ship in any iOS 26 release either. In March 2025 Apple said publicly that the work would "take us longer than we thought". In June 2025 Craig Federighi said it "needed more time to reach our high quality bar". In May 2026 Apple agreed to a $250 million settlement of a class action over the delay. The features finally appeared in beta with iOS 27 in July 2026.
Meanwhile, a different set of Apple Intelligence improvements did ship, starting with iOS 18.1 in October 2024. Those are real and they are what you have today. Both things are true, which is why the coverage has been so muddled.
What Works Right Now — With Apple's Own Examples
Recovering when you fumble a request
This is the most underrated shipped improvement. You no longer have to restart a request when you misspeak. Apple's own example:
Siri, set an alarm — wait no, sorry, I meant a timer for 10 minutes — actually, let's make that 15.
Old Siri would have set an alarm, or nothing. Current Siri follows the correction to the end and sets a 15-minute timer.
Context carried across requests
Follow-ups no longer need to restate the subject. Apple's example runs three deep:
How are the San Francisco Giants doing this season? — then — When are they playing next? — then — Add that to my calendar.
"They" and "that" both resolve from the conversation. This works today and it is the single most useful shipped capability.
Questions about your own devices
Siri absorbed Apple's support documentation, so it can answer how-to questions about iPhone itself rather than sending you to a web search. Apple's examples:
Where do I name a group text on iPhone?
How do I change what shows up in Control Center on my iPhone?
How do I play sound from my TV through HomePod?
This is quietly one of the best reasons to use Siri at all, and almost nobody knows it exists.
Type to Siri
Full capability, silently. On an Apple Intelligence iPhone you double-tap the bottom of the screen and type. It is more accurate than speech for unusual names and long requests, and it is the right tool in any room where talking to your phone is awkward.
ChatGPT hand-off
For open-ended questions Siri offers to pass the request to ChatGPT, with your permission. Useful, but be clear about what is happening: that is not Siri answering, and it is why answers to broad questions feel different in character from device questions.
Visual Intelligence — and what it is not
You can press the side button with volume up to capture what is on screen and ask about it. This is genuinely useful for identifying objects, translating text, or searching for something you can see.
But this is not the onscreen awareness Apple demoed. It is screenshot-triggered, you invoke it deliberately, and it routes to a search or chatbot backend. You cannot simply ask Siri mid-conversation about the thing you are looking at. Plenty of articles blur these two together; they are different features.
What Has Not Shipped Yet
These are the iOS 27 beta features. They are worth knowing about so you can recognise which reviews you are reading, but do not expect them on a current public release.
- Onscreen awareness — asking about what you are looking at without taking a screenshot. Apple's examples include asking where a photo was taken, or to translate a menu on screen.
- Personal context — resolving references against your own mail, messages and photos. Apple's examples: finding a flight confirmation number, or the lease a landlord emailed last month.
- In-app actions — completing multi-step tasks across apps, such as sending all of today's photos to a specific person, or adding a stop to a route already in progress.
Two caveats matter even once iOS 27 arrives. Apple's own wording is that Siri AI will be available as a beta, for devices set to English, with more languages to follow. And Apple's footnotes state it will not be available on iPhone or iPad in the EU at launch, nor in China. If you are reading enthusiastic coverage from a US reviewer, that context is doing a lot of work.
Which iPhones Support Any of This
Apple Intelligence requires iPhone 15 Pro, iPhone 15 Pro Max, or any iPhone 16 or iPhone 17 model. The base iPhone 15 and 15 Plus are not supported, which surprises people who bought a phone the same year. You also need roughly 7GB of free storage for the on-device models.
Some newer voice controls in the iOS 27 beta additionally require a device with more memory, which narrows them further within the supported list. That is a restriction on specific voice features, not on Siri AI as a whole.
What Is Powering It — Apple and Google
Worth knowing, because it explains a lot of the coverage you will see. Apple has confirmed on the record, repeatedly, that it worked with Google on the models behind Siri. Tim Cook said on an earnings call that Google's technology "would provide the most capable foundation" for Apple Foundation Models, and Craig Federighi said on stage at WWDC 2026 that Apple "partnered deeply with Google to leverage the technology behind the Gemini family of models".
What that does not mean is that you are talking to Google. Federighi was explicit that Apple uses none of the models Google deploys to its own customers. Apple's framing is that Gemini helped train Apple's models, with requests then handled on-device or on Apple's own Private Cloud Compute.
There is an unresolved tension here worth flagging honestly: Google executives have publicly described Google as Apple's "preferred cloud provider" for this work, which is not obviously the same story. Apple has not squarely addressed the discrepancy. Treat confident claims in either direction with suspicion.
Getting the Most From What You Actually Have
Three habits, all available today. Use follow-ups deliberately rather than restating full requests — that is the shipped strength. Ask Siri device how-to questions instead of searching the web, because that is a genuinely strong and undiscovered capability. And switch to typing when a spoken request fails twice, since the failure is almost always transcription rather than comprehension.
For the full walkthrough of activation methods, voice options and privacy controls, see our complete guide to Siri settings on iPhone. If Siri is not responding in your language, our guide to Siri language availability explains why the lists differ by feature.
Bottom Line
Judge Siri on what has shipped, not on the 2024 demo. Today it handles corrections, follow-ups, device questions and typed requests well — that is a real improvement and most people never use it. Onscreen awareness, personal context and in-app actions remain in beta, English-only, and unavailable in several major regions. Anyone telling you the rebuilt Siri is here is describing a beta.
Related iPhoneExtra Guides
More from iPhoneExtra: Siri 2.0 Complete Guide | Siri 2.0 Tips and Hidden Features | iOS 27 Beta 7: The Biggest Changes.
