AI in everyday apps
Which new AI features I actually use, which ones I've switched off, and how to tell the difference.
Which new AI features I actually use, which ones I've switched off, and how to tell the difference.
Somewhere in the last couple of years, every app on my phone got an AI button. My keyboard can rewrite my messages. My photos app can erase people. My notifications can summarise themselves. My search can explain things instead of just finding them.
Some of it is genuinely brilliant. Some of it is a sparkle icon on a feature nobody asked for. After living with all of it for a while, here's my honest sorting.
Removing things from photos. Google's Magic Eraser, Apple's Clean Up, Samsung's object eraser. Tap a stranger in the background or a bin next to a building, and it's gone. As a photographer I was sceptical, but for casual photos it saves a lot of shots. I'd never use it to pretend a scene was different from reality, but a random tourist's elbow? Gone.
Transcription and live captions. Voice memos and meeting recordings turned into searchable text, on the device. This is one of those quietly life-changing features. I record a thought while walking and find it later by searching a word.
Translation. Pointing a camera at a menu or a sign and reading it in my own language still feels like science fiction. Having lived and travelled abroad, this one has saved me more than once.
Search that understands what I'm looking at. Google's Circle to Search, and similar visual search features, let you circle something on screen and ask about it. Useful when you see a gadget in a video and want to know what it is.
The best AI features don't feel like AI. They just feel like the app got smarter.
Notification summaries, for important things. The idea is great: a stack of messages squeezed into one line. In practice, summaries sometimes miss the point or get it wrong, and that's not okay when it's news or something that matters. Apple even paused summaries for news apps in early 2025 after they produced misleading headlines. I keep them for group chats and turn them off for anything serious.
Rewriting my messages. It's impressive that my keyboard can make a text sound more "professional" or "friendly". But when it does, it stops sounding like me. For a work email, maybe. For a message to a friend? No thanks.
Chatbots squeezed into apps that didn't need them. I don't need to "chat" with my shopping app. I need it to find the thing and show me the price.
I've ended up with a simple test. A good AI feature does one of three things: it saves me real time, it does something I genuinely couldn't do before, or it removes a small annoyance without me thinking about it. A bad one makes me check its work, changes my voice, or adds a step to something that was already easy.
And one more rule: if it's going to read my messages, photos or voice, I want to know where that happens. A lot of new features run on the device itself, which I much prefer. When something gets sent to a server, I'd like to be asked.
I think the hype will settle down, the sparkle icons will slowly disappear, and the useful features will just become normal parts of every app. Nobody calls autocorrect "AI" anymore. In a few years, that's probably how we'll talk about erasing objects from photos or translating a menu.
My rule: use the AI that gives you time back. Switch off the AI that asks for your attention.
— Desly