For about two years, using AI meant opening a tab. You went to ChatGPT. You went to Claude. You went to Gemini. You typed into a box, waited, copied the answer, came back to whatever you were actually doing.

That phase is quietly transforming, and most people haven’t noticed because there was no announcement. AI is no longer a destination. It’s becoming a layer that sits on top of things you already use, activated by a small button you didn’t see last month.

Here are some of the small ones worth knowing about.

Google Search’s AI Overviews. The summary that now appears above your search results is doing the reading for you. For a simple factual question, you often don’t click through to any website at all. For a more tangled one (“best pilates studio near Indiranagar with a trial pack”), it will reason across several searches at once and give you one answer. Useful. Also quietly reshaping how the web gets read and who gets traffic.

AI Mode in Search. A separate tab next to “All” and “Images,” where the search engine behaves like a conversation. Ask a messy, multi-part question in one go. Follow up. The links still appear on the side, but the interaction is no longer ten blue links.

Ask Maps. Google Maps now takes conversational questions. “Where can I charge my phone without a long wait for coffee” is a real example Google uses. You’re not searching for a type of place anymore, you’re describing a situation and letting the map figure out the category. It can even book a reservation while you’re walking there.

Ask on YouTube. Below select videos, between Share and Download, there’s now an Ask button with a Gemini icon. You can ask what a video said about a specific topic, get a summary without watching, or take a quiz on the content. For long tutorials and lectures, it collapses an hour into a paragraph.

Ask Photos. Google Photos used to be a search engine for your own memory. Now it’s a conversation. “Show me the best photo from every national park I’ve visited.” “Remove the cars in the background.” You describe the edit, it finds the tool. The sliders are still there, but you don’t need to know they exist.

Gmail’s AI Overviews. Long threads get summarised at the top. Suggested replies draft a response in your rough voice. Help Me Write will build a first draft from a one-line brief. Useful when the inbox is a to-do list disguised as communication.

Everything else that doesn’t call itself AI. Spotify’s Discover Weekly. Netflix’s row order. Amazon’s “customers also bought.” Google Translate reading a signboard through your camera. Background removal in Canva. Live Caption generating subtitles on any video playing on your phone. Your photo app picking a face from ten thousand images. None of this announces itself. None of it asks you to open a separate app. It just works, and has been working for a while, and we stopped calling it AI because it stopped being new.

The move from destination to layer is the same move that electricity made, or search, or GPS. First it’s a thing you go to. Then it’s infrastructure, and it mostly shows up as a button you didn’t have to press yesterday.

There’s something worth naming about this shift. When AI was a destination, you were the one deciding to use it. You opened the tab. You framed the question. You evaluated the answer against your own reasoning and then decided what to do. The chat interface, for all its flaws, kept you in the driver’s seat because using it required a deliberate act.

The layer is different. The layer doesn’t wait to be asked. It summarises the thread before you decide whether you wanted a summary. It picks the top result before you see the others. It describes the video before you watch it. Each individual nudge is tiny. The cumulative effect is that a lot of small judgements, what matters in this email, which three reviews to read, whether to watch the whole thing, are being pre-made for you by a system whose reasoning you can’t inspect.

This is not a warning. It’s an observation. The tools are genuinely useful, and I use most of them. Ask Maps found a quiet cafe with a power socket. AI Overviews has saved me from reading six mediocre blog posts to confirm one fact. The Ask button on YouTube has rescued me from tutorials padded to hit the ten-minute watch-time threshold.

But the skill worth keeping is the one the layer makes optional: knowing when to switch it off. Reading the actual article when the stakes are real. Watching the full video when the nuance matters. Looking at the map yourself when you want to notice what’s around you, not just what the system thinks you were asking for.

The destination era taught us to think of AI as a separate thing we could engage with or ignore. The layer era asks a harder question. Not whether you use AI, you already do, dozens of times a day, but whether you still notice when you are.