Turning off the AI you didn't ask for

The first four chapters are about adding AI on purpose. This one is about removing the AI that showed up without asking — in your browser, your operating system, your phone, and the programs you were already paying for.

Why this belongs on a site about using AI

It might look like a contradiction. It isn't. Everything on this site rests on one idea: you decide what runs, and you decide what sees your information. A feature that switched itself on during an update is the opposite of that, no matter how useful it might be.

So this chapter isn't anti-AI. It's the same skill as the rest of the site, pointed the other direction. Turn things off, see what you actually miss, then turn back on the few things that earned it. That order matters — it's much easier to notice what a feature does for you when you've lived a week without it.

The two kinds of off switch

Some toggles hide a button. Others stop the processing. They look identical in a settings menu, and only the second one changes where your information goes. When a program offers both a long list of individual feature switches and one master switch, the master switch is almost always the honest one — reach for that first, then re-enable individual features if you find you want them.

Where these settings tend to live

Menus get rearranged constantly, so it's more useful to learn the pattern than to memorize a path. In nearly every case, the controls sit in one of three places:

Privacy settings — the most common home for anything that involves reading your content.
A section named for the feature itself — increasingly its own menu entry, often labeled "AI," "Assistant," or a product name.
Your online account, not the program — a frustrating but common pattern. The switch that governs whether your data is used for training frequently lives on a website, not in the software on your desk.

Search the settings for "AI," then for the assistant's brand name, then for "suggestions." Those three searches find most of it.

A sweep you can do in half an hour

Start with your browser. It sees more of your life than any other program on your computer, so it's the highest-value stop. As of early 2026, Firefox has a single master control that switches off every AI feature including ones added in future updates, and Edge has a comparable one-toggle setting. Chrome's controls sit under its privacy settings, split across several entries.
Then your operating system. On Windows, the screen-recording feature called Recall lives under privacy settings and has both an off switch and a button to delete anything it already captured — use both. On a Mac or a phone, look for the assistant's own settings panel and for anything offering to summarize your messages or notifications.
Then your phone. Same idea, smaller screen. Pay particular attention to keyboard apps and anything that offers to write replies for you, since those see everything you type.
Then the two or three programs you actually live in. Your email, your notes app, whatever you write in. Skip the rest — this is the step where people burn an afternoon for no benefit.
Write down what you changed. A short note in a text file. Updates sometimes switch these back on, and in six months you will not remember what the original state was.

"On-device" is better, but it still isn't your decision

More of these features now run on your own machine rather than a company's server, which is a real privacy improvement and worth acknowledging. But on-device processing still means a model was downloaded to your computer — sometimes several gigabytes of it — and installed because a company decided you should have it. If you want a model on your machine, Chapter 02 shows you how to choose one yourself. That's a different thing from finding one already there.

What isn't worth turning off

Restraint matters here, or this becomes a hobby instead of a half hour. Spell check, autocorrect, photo search, spam filtering, and voice transcription have all quietly used machine learning for years, they mostly run on your device, and they earn their keep. The features worth your attention are the newer ones that read your content in bulk — summarizing your inbox, watching your screen, or acting on your behalf.

That last category deserves the most caution. A feature that only answers questions is a tool. A feature that clicks, buys, or sends on your behalf is something else, and it's worth leaving off until you have a specific reason to want it.

When it turns itself back on

It happens, usually after a major update. This isn't necessarily sinister — settings genuinely do get reset when a feature is rebuilt — but the effect is the same either way. Check your list twice a year, or any time an update announces something new and exciting. Ten minutes, twice a year, is the actual maintenance cost of staying in charge of this.

One honest caveat about this chapter: the specific menu names above will drift, because they always do. The pattern won't. Privacy settings, the feature's own menu, and your online account — that's where these live, and that will still be true long after the labels change.