AI in your pocket: mobile apps and voice

A lot of people assume the phone app is the same AI with a smaller screen. That undersells it. On a phone two of the AI's senses come alive that barely get used at a desk: voice and the camera. Those are what make the phone a strength of its own rather than a shrunken copy of the site.

Voice mode: not dictation, a conversation

Voice mode (an actual spoken conversation rather than typing) is at its best in ChatGPT: it answers in a live voice with barely a pause, you can cut it off mid-sentence and it picks up. Rehearsing an interview on a walk, practising a foreign language, asking a question with your hands full in the kitchen — you already tried this in the exercise from the ChatGPT lesson. Gemini does the same and also sees your phone's context; Claude bets less on voice — text is its element.

The camera: an AI that sees

Point the camera and the model talks about what it sees: the ingredients on a package, a warning light on the dashboard, a blueprint, a screen with a cryptic message, a landmark. Photograph a receipt — "sort this into categories"; show it a vegetable bed — "what's wrong with these leaves?" At a computer you'd be describing all that in words; on a phone you just show it. That doesn't change convenience, it changes the range of things you can ask.

Gemini on Android: instead of the Google button

On Android, Gemini took the classic Google voice assistant's place in 2025: the same shortcut, but instead of "set a timer" you get a real conversation that can see the email on your screen and reach into your calendar and mail. For an Android owner it's the shortest path to AI — it's already in the phone, nothing to install.

iPhone: here the choice is yours

On iPhone the built-in helper is a weaker conversationalist, so you install the big three's apps by hand from the App Store — and that's arguably a plus: you pick whose voice and character suit you rather than settling for what came preinstalled. All three apps are free to start.

Claude on a phone: the strict writer

The Claude app is minimal and fast, built around text and documents: drop in a PDF straight from a messenger and it breaks it down on the move. Voice and camera aren't the point here. If your mobile scenario is "read and rewrite text from my phone", Claude fits; if it's "talk and show", take ChatGPT or Gemini.

The quiet conveniences you notice late

There are small things that make you open the app more often than the site. The share button: from any app — an article, an email, a note — you can send text straight to an AI without copying by hand: "summarise", "translate", "reply politely". Dictation on the move: mumble a messy thought into your phone in a queue and the model turns it into a tidy note. Answers out loud: a long answer you can listen to while walking instead of reading. Each of these saves ten seconds, but they add up over a day — which is exactly why the phone, not the browser, tends to become an ordinary person's way into AI.

When the phone wins and when the computer does

The split is simple, and everyone arrives at it by experience.

• Phone: talking out loud, "let me show you", a quick question in a queue, photographing a document on the spot.
• Computer: long text, lots of files and tabs, copying code, work where typing speed and a big screen matter.

One account — your conversation history syncs: start by voice on the road, open the same chat at home on the computer and finish it. These aren't two tools, they're one tool in two roles.

Common myths

Three stubborn myths surround the phone, and they keep people from installing the app for years.

• "The mobile version is cut down." The opposite: the phone has things the desktop doesn't — live voice and a camera.
• "Voice is just for dictating." Modern voice mode is a conversation with intonation that you can interrupt, not a tape recorder.
• "AI doesn't work on iPhone." It does — you just install it as a separate app, and the choice of model is yours.

Do this now

Two minutes, for real. Open any all-rounder's app on your phone, turn on voice mode and ask a question out loud — say, have it explain a story from today's news. Then point the camera at anything with dense text on it (a package, an instruction sheet, a bill) and ask it to break it down. You'll feel exactly what no review screenshot conveys.