
Apple Intelligence might be the headline feature of the iPhone 16 marketing campaign, but in many cases, you might not even notice it's there.
Apple's take on generative AI is out today for Australian iPhone users (with an iPhone 15 Pro or iPhone 16) as part of iOS 18.2, and on Mac and iPad. I've been testing the feature set for a few weeks now across my iPhone and Mac, and it comes across as a mix of playing catch-up and low-stakes features.
As a journalist, the Apple Intelligence feature I've found most useful is transcription in voice memos and phone calls. This has been available on Pixel devices for years, and as with Google's take, it isn't perfect, but it's good starting point and a genuine time saver.
There's also an image clean-up tool similar to Google's magic eraser. This is used to remove unwanted details in photos, and it does an okay job! Results can vary a lot depending on the complexity of your photo, however. While they're certainly passable at a glance, textures like grass aren't quite right if you're looking closely.
iPhone 16 owners also get Visual Intelligence, a Google Lens-type feature you can activate via Camera Control. With it, you can either ask ChatGPT or Google about whatever you're seeing in your camera viewfinder. Given you could already use Google Lens on iPhone, Visual Intelligence isn't a huge deal. Apple isn't doing anything different here, it's just giving you a quicker way to access Google's tech.
A large number of Apple Intelligence features are built around summaries. There are notification summaries, which attempt to give the gist of what's happening if you have multiple notifications from a single app. These can be too vague to helpful if you've got a group chat popping off, but notifications summaries are mostly inoffensive and sometimes hilarious.
Take this example from a group chat discussing the whole martial law thing in South Korea:

Apple's default Mail app will also give you a short summary of new messages below the subject. It does a good job of pulling out information like tracking numbers, but there's no way to engage with them without opening the email. While the extra step isn't a massive time sink by any means, the current execution feels like it's halfway there.
There are also Summarise buttons in the Mail app and Safari. These condense an email or a webpage down to a two- or three-sentence block. There are no options to get a longer summary, as with Samsung's Galaxy AI, for example. The shorter summaries seem to reduce the risk of errors, but I've still had the feature lie. When summarising my iPad mini review, it consistently made up information about the device having a higher price point, despite that not being mentioned in the story (or being true).
Other generative AI features are the kind of table stakes inclusions you'd expect to see. There are Writing Tools with the typical rephrase, proofread, and summarise functionality that's par for the course. I'm not the biggest fan of this type of feature - partially because writing is my livelihood, and partially because using AI to write stuff you expect other people to read feels rude.
Similarly, you've got basic image generation via Image Playground and Genmoji. I find the Image Playground aesthetic unsettling and it refuses to acknowledge that I have long hair. The overall app is mostly harmless though, and it seems to have pretty strict guardrails – you're not going to be infringing on Nintendo's copyright any time soon. Genmoji is even more basic, and is solely used for generating custom emojis. They're fine?

What's interesting is how out of the way these features are. There aren't buttons you'll accidentally tap that fire up features when you don't want them to, as in Meta's apps and the latest Gmail refresh. A fellow tech journalist didn't realise that Genmoji was part of the beta, as the icons you tap to fire it up isn't exactly prominent. It's the same with summary features, where you first need to open reader mode in Safari to find the button, or scroll up in an email.
It's also worth noting that a lot of this functionality happens on device, and Apple has taken its usual stance on privacy with generative AI. None of your data is used for training, nothing is linked to you, and Apple will even anonymise data sent to ChatGPT if you're using it via Siri.
In some ways, restraint is the story when it comes to Apple's take on generative AI. While Apple Intelligence is the headline feature splashed across every marketing image, these aren't the kind of features that sell a phone. They're nice to have, but hardly game-changing. Apple Intelligence doesn't have the same kind of scope as Google's Gemini, for example. (Which I'd also argue doesn't sell a phone.)
There's a larger Apple Intelligence roadmap with more ambitious features coming next year, but for now, there are two ways to look at this.
The simple explanation is Apple is behind on generative AI and playing catch up. Apple didn't anticipate just how big generative AI would get and now it can't afford not to invest in the tech.
The more interesting take is that Apple is being cautious. There are lots of issues with generative AI from both philosophical and practical perspectives. On one hand, you've got legal, ethical, and environmental concerns, and on the other, you've got functionality. Even if we just ignore the wider picture and look at generative AI from a pure product perspective, it often doesn't work how you'd expect.
The core problem is generative AI doesn't actually know anything. It's just autocorrect on steroids, and it consistently makes things up. That's a problem the industry is no closer to solving, and it results in a poor user experience.
When Google rolled out generative AI to search earlier this year, there were countless pieces and posts mocking it for suggestions like putting glue on a pizza. Even if we're looking at generative AI as a feature on phones, it's inconsistent at best. Gemini consistently failed at tasks that Google suggested it could do when I was reviewing the Pixel 9. It's simply not reliable.
Through this lens, it makes sense that Apple is starting its generative AI journey with lower stakes functionality. It helps prevent these kinds of issues, and it always ends up being a much bigger deal when Apple gets something wrong. Apple Maps still has a reputation problem following its bungled launch over a decade ago, despite being very good now.
Apple Intelligence is a juggling act. It has to placate investors who believe Apple isn't doing enough in generative AI, it has to mesh with Apple's positions on privacy and the environment, and it can't cause Apple reputational damage.
Those are tricky priorities to balance, especially given the issues with generative AI, including hallucinations and a lack of clear use cases. Generating a silly image via a prompt is hardly the future we were promised.
Generative AI features are novel but unreliable. Sometimes they can be helpful, but they're far from revolutionary. Other phones may have more generative AI features than the iPhone, but at the end of the day, they're just features – not a killer app.
In a roundabout way, Apple's low-stakes implementation of generative AI could be the ideal position.
Sign up and get money-saving deals, telco news and more.
These are the most popular NBN 50 plans with WhistleOut users this week: