Ask Siri to compare two PDFs you just downloaded and it will cheerfully offer you a web search. That gap between what Apple’s assistant does and what people actually want is why so many iPhone owners went looking elsewhere and ended up installing Gemini. The app has been in the App Store since late 2024, it costs nothing to try, and it handles things Siri still can’t touch: reading long documents, holding a spoken conversation you can interrupt mid-sentence, and looking through your camera at whatever you point it at.
Here’s what Gemini on iPhone is genuinely good at, where it stumbles, and the settings to change before your first real chat.
Gemini on iPhone is a separate app, not a Siri upgrade
This trips people up constantly. The Gemini app is a standalone download. You tap its icon, you talk to it there, and everything stays inside that app. It cannot set an alarm, send a message in Messages, or turn off your living room lights, because iOS doesn’t hand third-party assistants that kind of system access.
What it does have is a direct line into your Google account. Connect it and it can read Gmail, check your Google Calendar, and pull notes out of Keep. If you live inside Google’s apps but carry an iPhone, that combination is quietly more useful than most of what Apple Intelligence has shipped so far.
Setting it up takes about two minutes
Download from the App Store, sign in with a Google account, and allow microphone and camera access when the prompts appear. Grant the camera permission. A large chunk of what makes Gemini feel different on a phone only works when it can see, and the app is useless for those features if you’ve denied it.
To get it within thumb reach, build a Shortcut that opens the app and assign it to the Action Button. It isn’t the same as a proper system assistant, but it’s two seconds faster than digging through folders.
Free vs paid: what you actually get
- Free tier: Gemini’s faster model, voice conversations, image generation, photo and file uploads, and a small number of Deep Research runs per month.
- Google AI Pro (roughly $20 a month): the stronger Pro model, much higher limits on uploads and Deep Research, longer context, and extra storage across Google’s apps.
- Students and promotions: Google has run free or discounted AI Pro windows in several countries. Worth checking before you hand over a card.
Start free. Upgrade the month you find yourself hitting an upload limit twice in a week, not before.
Gemini Live is the reason it stays on your home screen
Talking to a chatbot sounds like a gimmick until you try it with your hands full. Gemini Live keeps a real conversation going, lets you cut in halfway through its answer, and doesn’t require you to phrase things like a search query. You can switch to the camera mid-chat and point it at whatever is confusing you: a router with one blinking amber light, a menu in a language you don’t read, the care label on a jumper you’re trying not to shrink.
Latency is the thing to watch. On decent Wi-Fi it feels close to a phone call. On a weak cellular signal it turns into a walkie-talkie. Better camera hardware helps a little here, though even the sensor upgrade in our iPhone 18 Pro review won’t rescue a badly lit room.
The difference between a chatbot that technically runs on a phone and one designed to be used one-handed is bigger than most people expect, and it’s covered well in what actually works when you’re holding a phone.
Long documents, photos and the trick Siri can’t do
This is where the gap is widest. Drop a 40-page tenancy agreement into Gemini and ask which clauses mention break fees. Upload a photo of a handwritten receipt and ask it to total the items. Paste a spreadsheet export and ask which three customers spent the most last quarter. It handles all of it, and the free tier gives you enough room to know whether you’d pay for more.
Two habits make a real difference. Rename files descriptively before uploading, because the app uses the filename as context. And ask for a specific shape of answer: “a table with three columns and no commentary” gets you something you can use, while “summarise this” gets you mush.
Gemini vs ChatGPT on iOS
Both apps are polished, both handle voice, and both will happily write your awkward email. Gemini’s edge is the Google account sync and the camera work. ChatGPT’s is a broader set of connectors and, for a lot of people, slightly better long-form writing. We went through the trade-offs properly in Gemini vs ChatGPT: where each one actually wins, and the short version is that your existing digital life decides it for you. Gmail, Calendar and Drive? Gemini is the obvious pick. Everything else spread across other tools? Less clear-cut.
The Siri deal that will change all of this
Apple has reportedly been working to squeeze a very large Gemini model down to something an iPhone can run, as a way of rebuilding Siri from the inside rather than bolting on a chatbot. The details are worth reading in Apple working to cram massive Gemini model into iPhone to power new Siri.
If that lands as described, the standalone app stops being a workaround and starts being a companion. Treat any timeline with suspicion, though. Apple has missed self-imposed deadlines on assistant features before, and shipping a model of that size without melting the battery is genuinely hard.
What Gemini still can’t do on iOS
- System actions. No alarms, no timers through the Clock app, no sending texts as you. Anything that needs an iOS API stays out of reach.
- Background work. Close the app and it stops thinking. Nothing is running in the background waiting to surface something.
- Deep widget support. The home screen widget is functional rather than clever. You can’t fire a prompt straight from it the way you’d hope.
Google is edging toward a more proactive assistant elsewhere, and it’s further along than you’d guess. Our write-up on Google’s 24/7 Gemini Spark assistant shows what happens when the model gets to start conversations instead of waiting for them. None of that proactive behaviour is on the iPhone app yet.
Three privacy settings to change on day one
Gemini chats are saved to your Google account by default, and Google’s standard retention window is 18 months. That’s a long time for the questions you ask at 11pm.
Shorten the auto-delete window
In your Google account settings, find Gemini Apps Activity and change auto-delete from 18 months to 3 months or, if you’d rather, 72 hours. You can also pause the activity entirely, which means new conversations aren’t stored in your account.
Turn off human review
There’s a separate toggle for letting reviewers read excerpts of your chats to improve the model. Flip it off unless you’re genuinely fine with that.
Think before connecting Workspace
The Gmail and Calendar extension is the best feature in the app and also the one with the most reach. Once it’s on, your prompts can pull in real email content. Leave it disconnected until you actually need it, and disconnect when you’re done testing.
Where it earns a spot on your dock
The honest test is whether you open it unprompted in week three. For me it’s three recurring jobs: a 7:40am “what’s on today and what did I agree to in email yesterday”, the mid-week recipe question that used to take four separate searches, and anything involving a photograph of text I can’t be bothered to read properly. Those alone justify the icon sitting next to Messages.
What it won’t do is replace the assistant that lives in your phone’s bones. Not yet. Use Gemini for the thinking work and accept that Siri still owns the alarm clock, and the two of them get along fine.