Blog

Apple's Built-In Predictions: Good, Patchy, and Not a Habit

Let me start by giving Apple credit, because it's due. If you've typed on an iPhone, you've used predictive text, the little gray words above the keyboard that guess what you're about to type next. It has been there for years, it's genuinely decent, and it taught a billion people that software can finish your words. That is a real achievement, and I'm not going to pretend otherwise.

Here is the thing, though. On the Mac, the same feature is patchy in a way that matters, and it never quite becomes a habit, because a habit needs to work everywhere, every time, with the same keypress. That is the gap I built WriteAmp to fill, and I want to explain it honestly, including the parts where Apple's version is genuinely fine.

What Apple's predictions actually are

Apple's text prediction works on the iPhone keyboard, and on the Mac it shows up as inline predictions in some Apple apps and in some fields. Type part of a word or a sentence, and the system offers a completion. It is free, it's built in, and in the apps where it appears it's usually correct enough to be useful.

The key phrase is "in the apps where it appears," because that's the whole story. Apple's predictions aren't system-wide. They show up in Apple's own apps and in certain fields, and they don't appear everywhere you type. You can be typing along in Mail and get a helpful prediction, then switch to a third-party app, or a browser, or a chat field, and the predictions are just gone. The feature exists, but it's not a layer under everything, the way the Accessibility system is.

The habit problem

Here is the part that's subtler than coverage. A feature you can use isn't the same as a habit you can build, and habits need consistency. If predictions show up in some apps but not others, your brain never learns to reach for them, because it can't predict when they will be there. You type in Mail and the gray words appear, and you start to trust them, and then you switch to your browser and they're gone, and your brain quietly files the whole thing under "unreliable" and stops reaching.

That is the real gap. Apple's predictions aren't bad, they're inconsistent, and inconsistency is what kills habits. I wrote about how the reflex forms in another post, and the whole argument there depends on the tool being there every time. A reflex can't form around a feature that appears and disappears based on which app you happen to be in.

What Apple gets right

I want to be fair about the other direction too. Apple's approach has real advantages that a third-party tool can't match. It is free, which is hard to argue with. It is deeply integrated, which means it works in Apple's apps with no permissions and no setup. And on the newer Macs, Apple has been pushing more capable on-device writing models, which closes some of the quality gap. If you mostly type in Apple's own apps, the built-in predictions may genuinely be enough for you, and I wouldn't try to talk you out of them.

There is also a point worth making plainly. Apple's predictions and a tool like WriteAmp aren't enemies. They can overlap, and in practice the system-level stuff and a third-party autocomplete coexist, because they trigger on different things. The honest framing isn't "Apple is bad, buy mine." It is "Apple is good in its lane, and its lane doesn't cover everywhere you type."

How to decide

Here is the practical question to ask yourself.

The privacy point is worth a moment, because it's easy to assume Apple's free built-in is the private option and every third-party tool is a data collector. That isn't how this category works. There are cloud autocomplete tools that send your typing to a server, and there are local ones like WriteAmp that don't. Apple's predictions are local. The choice between Apple's and WriteAmp isn't a privacy choice, it's a coverage and habit choice. I wrote about the network ledger so you can see exactly what a local third-party tool does and doesn't send.

The honest close

Apple built a genuinely useful thing, and then left it patchy, and the patchiness isn't an accident you can fix by wishing. It is the shape of a platform feature that has to serve Apple's apps first. A tool built for one job, system-wide autocomplete, can afford to be obsessive about the other apps in a way a platform feature can't.

If the built-in predictions cover you, use them, free and happy. If you've hit the wall where the gray words vanish the moment you leave Apple's apps, the trial is free for 30 days and it's the same gray words, everywhere, with one key. The compare page has the fuller picture of where everything fits, and the map post shows the whole landscape if you want to see all the options at once.

---

On macOS 26+ Macs, Apple Intelligence mode stays free even after the trial ends — you always keep a working path to suggestions.

Written by Amit Ashwini, who builds WriteAmp and runs its marketing. More: why the Tab key beats the chat box · mini, midi, and max compared · benchmark methodology.