Predictive text is a keyboard feature that suggests your next word, or a full phrase, as you type. It shows up across smartphones, computers, and apps like Gmail and Google Docs, and its job is to help you type faster with fewer errors. Tap a suggestion, keep moving.
It’s become so common that most people use it without thinking about it. But there’s a lot going on under the hood, and knowing how to manage predictive text well, and when to reach for something more powerful, can make a real difference in how fast you communicate at work.
What is predictive text?
Predictive text is a system that analyzes what you’ve typed and surfaces likely next words based on language patterns, context, and your personal writing habits. The feature shows up as a row of suggested words above your mobile keyboard, or as grayed-out text completing your sentence in apps like Gmail.
The technology started with T9, the number-pad input system on flip phones that guessed which word you meant from the letters on each key. Clunky, but the underlying idea was the same as today. Modern predictive text keyboards have come a long way: they learn from your messages, factor in the app you’re using, and suggest whole phrases at once.
The term “predictive text keyboard” covers a range of features, from basic next-word suggestions on a phone to full-sentence Smart Compose in Google Docs. The core technology is the same. Implementation varies by platform.
How does predictive text work?
Simple version: the system looks at the words you’ve typed and calculates which word is most likely to come next, based on patterns learned from a large amount of text.
The longer version involves three generations of technology.
N-gram models were the earliest approach. An n-gram model looks at the last N words you typed and picks the statistically most common word to follow them. Type “Happy” and a bigram model suggests “birthday” because that pair appears often in text. Surprisingly effective for something so simple.
Machine learning moved the needle significantly. Instead of fixed pattern tables, ML models learned from user behavior over time, picking up individual habits, adapting to slang and domain-specific language, getting better the more you used them. This is when phones started feeling like they actually knew you.
On-device neural language models are where things stand today. Modern predictive text in iOS 17 and later, Gboard, and SwiftKey runs transformer-based models directly on your device. It’s the same basic architecture behind large language models, scaled down to fit on a phone. That matters for two reasons: it’s faster, and your data stays on your phone. These models understand context across a longer stretch of text, which is why suggestions sometimes feel eerily accurate.
The personalization layer sits on top of all of this. Your keyboard tracks your vocabulary, the contacts you message most, and even the apps you’re in when you type. You write differently in Slack than in a birthday text.
How to use predictive text on your device
Android
To enable predictive text:
- Open Settings
- Tap System > Language & input > On-screen keyboard
- Select your keyboard (Gboard or Samsung Keyboard)
- Tap Text correction
- Toggle Show suggestion strip or Next-word suggestions on
To reset or retrain suggestions: In Gboard, go to Settings > Dictionary > Delete learned words and data to clear personalized suggestions and start fresh. Useful if suggestions have picked up terms from a sensitive conversation or you’ve switched between professional and personal use on the same device.
Practical tip: If Gboard is suggesting terms from the wrong context, add a secondary language under Languages. Suggestion pools are partially separated by language, which can reduce bleedover between work and personal vocabulary.
iPhone
To enable predictive text (iOS 17 and later):
- Open Settings
- Tap General > Keyboard
- Toggle Predictive Text on
- Also enable Show Inline Predictions for full-sentence completions that appear directly in text fields
To reset or retrain suggestions: Go to Settings > General > Transfer or Reset iPhone > Reset > Reset Keyboard Dictionary. This clears all learned vocabulary. Worth doing if autocorrect has accumulated too many bad habits.
Practical tip: Tap and hold the space bar to turn the iOS keyboard into a trackpad. Drag left or right to move the cursor precisely. It’s the most underused iPhone keyboard feature and saves real time when you need to edit mid-sentence.
macOS
To enable inline predictive text (macOS Sonoma and later):
- Open System Settings
- Click Keyboard
- Under Text Input, click Edit
- Toggle Show inline predictive text on
Predictions appear as light gray text after your cursor. Press the right arrow key to accept, or keep typing to dismiss.
Practical tip: macOS predictions work best in native apps like Messages and Mail. Third-party apps may have limited support depending on how they handle text input.
Windows
To enable text suggestions:
- Open Settings
- Go to Time & language > Typing
- Toggle Show text suggestions as I type on
Practical tip: Text suggestions work in most system-wide text fields on Windows 11. If they don’t appear in a specific app, that app may be overriding the OS’s native text handling.
New Outlook and Microsoft 365
Microsoft Editor, available in Word, Outlook, and Teams, provides two types of predictive-style assistance: text predictions (inline completions accepted with Tab) and, in supported apps, Copilot-powered suggestions. Text predictions are on by default in the new Outlook. To manage them, go to Settings > Mail > Compose and reply and look for Text predictions or Writing assistance under composing options.
How to use predictive text in popular apps
A few apps go beyond keyboard-level suggestions and build prediction features directly into their interface:
Google Docs: Smart Compose is on by default. Start a sentence and gray suggested text appears after your cursor. Press Tab or the right arrow key to accept. Turn it off under Tools > Preferences > Show Smart Compose suggestions.
Microsoft Word: Editor text predictions work similarly: suggestions appear inline as you type, accepted with Tab. Manage the feature under File > Options > Advanced > Editor Settings.
Messaging apps: WhatsApp, iMessage, and most messaging apps use your phone’s native keyboard suggestions. The experience depends on your keyboard app. If you want stronger predictions in messaging, switching to a more capable keyboard is the fastest improvement. See our roundup of the best predictive text apps to compare options.
Predictive text vs. text expansion: What’s the difference?
Short version: they’re not the same thing, and they’re not competing for the same job.
Predictive text is reactive. The system watches what you type and offers suggestions based on patterns. You’re still doing most of the typing. It’s filling in guesses. The output is probabilistic. Think of it as the keyboard’s best estimate of what you meant.
Text expansion is proactive. You define an abbreviation, assign it to a block of text, and the expansion happens the moment you type that abbreviation, in any app, on any platform. The output is exact. Every time.
Here’s where the difference plays out in practice:
| Feature | Predictive text | Text expansion (TextExpander) |
|---|---|---|
| Triggered by | Your typing patterns | A specific abbreviation you define |
| Output | Suggested word or phrase | Exact text you specified |
| Learns over time | Yes — adapts to your habits | Only if you update your Snippets |
| Works across apps | Keyboard-level (most apps) | System-wide (all apps) |
| Team sharing | No | Yes — shared Snippet libraries |
| Best for | Casual messages, short replies | Professional responses, templates, forms |
| Accuracy | 80–95% for common words | 100% — it’s exactly what you saved |
Predictive text is great for everyday typing: quick texts, common phrases, low-stakes communication. Text expansion is for when you need the exact right words, every time. Support responses, medical documentation, legal disclaimers, client emails. Contexts where “close enough” costs you.
The two also work at different scales. Predictive text is a word-ahead or phrase-ahead feature. Text expansion handles anything from a two-word phrase to a full multi-paragraph template with fill-in fields.
There’s also an autocomplete angle worth understanding. Autocomplete and predictive text are often used interchangeably, but they work differently at the technical level.
Your responses, exact every time.
Predictive text makes its best guess at your next word. TextExpander expands to exactly what you saved. The same response, for every team member, every time. No variation, no wrong guesses.
Free 30-day trial. No credit card required.
Go further with TextExpander
Predictive text reacts to what you’ve typed. TextExpander works differently: you define exactly what you want to type, assign it a short abbreviation, and it expands anywhere you type it, the same way, every time.
Think of the difference as reactive vs. proactive. Predictive text makes an educated guess about your next word. TextExpander executes the text you’ve already decided on.
Here’s what that looks like in practice:
Multi-line email reply
Abbreviation: ;reply
Expands to:
Hi [name],
Thanks for reaching out. I’ve reviewed your message and will put together a full
response with everything you need by [date].
In the meantime, feel free to reply here or reach me directly at [phone].
Best,
[Your name]The bracketed fields are fill-ins. TextExpander pauses and prompts you to complete each one before the expansion finishes. You get a structured starting point you can adapt in seconds, not a rigid canned reply.
Introduction with fill-in fields
Abbreviation: ;intro
Hi, I’m [fill: Your name] from the [fill: team] team. I wanted to reach out about [fill: topic] — let me know if you have a few minutes this week.Short, high-frequency Snippet
Abbreviation: ;ty
Thank you for reaching out. I’ll review your message and get back to you within one business day.That last one saves maybe 15 seconds per use. Multiply by 20 times a day and the math starts to matter.
The real advantage over predictive text is consistency. A Snippet expands to exactly the same text every time, for every team member who uses it. No variation, no wrong guesses. You can also use TextExpander for autocorrection to replace specific words or phrases you reliably mistype.
If you’re new to Snippets, we have a practical guide on creating and organizing your first Snippets.
Build your Snippet library in minutes.
Every template above takes just minutes to set up in TextExpander. Add fill-in fields for on-the-fly personalization, then share your Snippets with your whole team so everyone starts from the same place.
Free 30-day trial. No credit card required.
Tips for getting the most out of predictive text
A few things that make a real difference:
- Accept suggestions regularly. The more you tap suggestions, the more the system learns your patterns. Ignoring them consistently slows improvement.
- Delete unwanted predictions. On mobile, long-press a suggestion in the bar and drag it to the trash to remove it. Worth doing with typos or terms you don’t want suggested.
- Match your register. Predictive text doesn’t know you’ve switched from Slack to a client email to a medical form. Being deliberate about language in different contexts, and resetting suggestions when things go sideways, keeps quality up.
- Use both tools together. For casual communication, predictive text handles the filler. For anything professional, pair it with TextExpander Snippets that work like a spell-checker and more so the content that has to be right always is.
Conclusion
Predictive text does the boring part of typing better than it used to. From T9’s number-pad guesses to today’s on-device transformer models, the improvement has been real, and the feature is worth actually configuring rather than leaving on defaults.
But predictive text has a ceiling. It’s reactive: it can only guess at what comes next. For professional communication where words need to be exactly right, a tool like TextExpander gives you control that predictive text can’t provide.
Most people find the best approach is both. Predictive text for quick messages and casual typing. Text expansion for anything that matters.
Looking for stronger predictive features? Our guide to the best predictive text apps covers the top options across platforms. And if you’re ready to go beyond guessing entirely, see our roundup of the best text expander tools available today.
Give your whole team exact responses, every time.
Predictive text adapts to each person’s habits. TextExpander gives every team member the same approved responses, templates, and scripts all updated in one place, instantly available to everyone.
Trusted by 100,000+ professionals. Free for 30 days.
Frequently asked questions about predictive text
What is predictive text?
Predictive text is a feature built into smartphones, computers, and many apps that suggests words or phrases as you type, based on language patterns and your personal writing habits. It appears as a suggestion bar above mobile keyboards or as inline gray text in apps like Gmail and Word. The goal is to speed up typing and reduce errors by offering likely next words before you finish typing them.
How do I turn on predictive text on iPhone?
Open Settings > General > Keyboard and toggle Predictive Text on. On iOS 17 and later, you can also enable Show Inline Predictions for full-sentence completions that appear directly in text fields. To reset your keyboard’s learned vocabulary, go to Settings > General > Transfer or Reset iPhone > Reset > Reset Keyboard Dictionary.
How do I turn on predictive text on Android?
Open Settings, then go to System > Language & input > On-screen keyboard, select your keyboard (Gboard on most devices), and tap Text correction to enable Show suggestion strip or Next-word suggestions. The exact path varies slightly depending on your device manufacturer and Android version.
Is predictive text the same as autocorrect?
No. Autocorrect automatically replaces a word it thinks is a typo. It changes teh to the without any input from you. Predictive text suggests words you haven’t finished typing yet and only activates if you tap or accept the suggestion. Both features typically run on the same keyboard, which is why they get confused.
Can I use predictive text on a desktop?
Yes. macOS supports inline predictive text in Sonoma and later, enabled under System Settings > Keyboard > Text Input. Windows offers text suggestions in many system-wide text fields. Microsoft Word and Google Docs both have built-in prediction features that work on any computer. Word uses Editor, Docs uses Smart Compose.
What is the difference between predictive text and text expansion?
Predictive text suggests what you’ll probably type next, based on learned patterns. Text expansion replaces a short abbreviation you define with a pre-written block of text: one sentence, a full template, or anything in between, with fill-in fields where needed. Predictive text is reactive and probabilistic. Text expansion is proactive and exact. They work well together: predictive text handles casual typing, and TextExpander handles professional content that needs to be consistent every single time.
