Apple has unveiled its Siri AI update β which will let your iPhoneβs digital assistant search the web, sort through messages, look at photos, and access a bunch of apps to complete your requests β as part of this Septemberβs iOS 27 software.
The announcement might not have done much to impress analysts and investors, who have been on Appleβs case to βcatch upβ with Google in generative AI for years. But regular iPhone users may be pleased that Apple has stuck to its usual modus operandi, turning chaotic tech into a polished utility, rather than moving fast and breaking things.
Appleβs own AI models are built on a foundation of Google technology, but the approach the companies take to integration is very different. As Google finds new ways to hoard the photos, text and audio you use to prompt Gemini to retrain it, or even feed the music that creators publish on YouTube into AI that can replicate it, Apple has promised not to store or look at anything. While Google adds new paid tiers that let Gemini act as a health coach or security guard, Apple is more measured and controlled. And free, for now.
There remain some fundamental ethical issues with generative AI that have gone largely unaddressed, and Appleβs approach merely stacks them neatly by the back door where they donβt need to be unpacked, which might rankle some of its creative users who take issue with the premise of the technology at all. But if you do want a capable digital assistant baked into your phone, and you understand that itβs likely to make some mistakes, the new Siri compares favourably to the similarly integrated experiences on Android phones.
Iβve been testing an early preview as part of the current iOS 27 developer beta. Itβs important to remember that features are subject to change between now and September, but itβs clear already that the updated assistant is significantly more helpful.
Head to head with Google
Comparing Siri AI and Gemini directly on several prompts, Siri performed somewhere between Geminiβs Flash and Thinking modes for speed and complexity, delivering quick, suitable answers. Both were good at identifying plants and animals, gave advice about household and work tasks, could search the web to summarise recent info for specific events, and digested files and big blocks of text to answer questions.
Interestingly, Siri was more curt and matter-of-fact, which I appreciated. For a prompt about advice on watching the World Cup as a casual football-enjoyer, Siri gave a brief and informative rundown while Gemini went for more than 1000 words with a party boy persona. (βWelcome to the hype train!β ). When given a list of requirements and asked to plan activities for a trip to Osaka, they both gave authoritative advice, but Siri was less indulgently evocative with its descriptions.
When I showed both assistants a forgotten security tag on a new pair of hiking boots, Gemini gave helpful advice against cutting it. Siri also said no, but more brusquely, implying that tampering with the security device on stolen shoes would constitute an additional crime. In this one example, Geminiβs friendly repartee was actually preferable because Siri came off like a cop.
If youβre not using the dedicated Siri app and just summoning the chatbot to ask about something you can see on your screen, it gives brief answers that I generally didnβt find very useful, but this is often true for Gemini too. Siri at least almost always verbalised its sources of information out loud, though that doesnβt mean much if itβs misidentified the thing itβs looking at.
Delving into your data
So what about personal context? I asked Siri if there were any school events coming up for either of my kids, to save me trawling back through weeks of emails. It found two immediately, both for my youngest child, and said there werenβt any for the older. I went back through my email to check, and it did miss a note about specific requirements for grade four homework next week. But I had asked for βeventsβ, so it may have made a judgement call on that. I followed up by asking it to put both events in my calendar, and it did. I asked if there were any other important recent messages, even if they werenβt tied to events, and Siri dutifully pulled up the homework reminder.
To try something more vague, I asked about t-shirt designs my wife had sent me months ago as an MMS. This time Siri didnβt do quite as well. It initially came up with nothing, but after a little back and forth it found the message and even reminded me of what I had replied at the time. I asked if it could find the specific shirts online, but I think it used my spoken context (rather than the images in the message) and only gave me generic places to buy t-shirts.
Searching through photos I had synced to iCloud was trivial for Siri. I asked where I found those tawny frogmouth owls last year, and it pulled up the photos and correctly reported the location data. It did misidentify a kookaburra as an owl, though.
Once corrected, it easily mapped directions to the right location in Maps. Clear requests that use a combination of apps also went well, including asking when I should leave for an event, telling Siri to find a recipe for vegan pancakes and add all the ingredients to my shopping list, or asking it to grab a document Iβd just received, check it for inaccuracies, and show me reputable sources of info if it found any.
Ultimately, while it may lack Geminiβs conversational flair and constant feature changes, Siriβs focus on privacy and utility may deliver what most iPhone users actually want.
Get news and reviews on technology, gadgets and gaming in our Technology newsletter every Friday. Sign up here.
Tim Biggs is a writer covering consumer technology, gadgets and video games.Connect via X or email.