Word prediction software for Mac in 2026: who it helps and which tools still run
Word prediction software speeds some writers up and slows others down. What ten studies show, which tools run on macOS in 2026 (Co:Writer is gone), and prices checked October 2026.
In 1994, word prediction made typists with high-level spinal cord injuries slower (Koester & Levine, 1994). In a 2014 study of people typing with a single switch, it took them from 1.51 to 2.72 words per minute — but only after the setup was retuned (Koester & Arthanat, 2017). Same feature, opposite outcomes.
Word prediction software pays off when each keystroke is expensive — a switch, an onscreen keyboard, a word you can't spell — and, in phone-keyboard studies, costs fluent typists some speed while still cutting their keystrokes. On a Mac in 2026: switch on Apple's free inline predictions or the Accessibility Keyboard's suggestion bar first; add WordQ or Ghotit if spelling is the problem, a system-wide autocomplete if tired hands are, and Voice Control if typing itself is the barrier. Co:Writer word prediction software, which several of the guides ranking for this search still list (checked October 2026), was retired on 31 December 2025.
I build TypeTab, one of the autocomplete apps in the table below. It gets one row and one section with its limits; everything else links to the study or vendor page it came from, with prices checked October 2026.
Which word prediction tool fits your need?
Pick by what a keystroke costs you, not by feature list.
| If this is you | What the studies found | Start with (free) | Paid tools worth a trial |
|---|---|---|---|
| You often get the start of a word wrong (dyslexia, dysgraphia) | Older prediction failed these writers because it needed the right first letters; prediction programs with phonetic matching raised word counts and fluency for six students (Evmenova et al., 2010) | Apple doesn't document sound-based prediction for either built-in option — use the WordQ 30-day trial | WordQ for macOS, Ghotit Real Writer & Reader |
| You spell fine, type on a keyboard, and your hands tire | In phone-keyboard studies, suggestions cut keystrokes but slowed typists down (Quinn & Zhai, 2016; Palin et al., 2019) | macOS inline predictive text | A system-wide autocomplete (TypeTab, Cotypist); for fixed phrases, a text expander (see text expander vs AI autocomplete) |
| You type on an onscreen keyboard, with a head pointer, eye gaze or a switch | 7 of 10 onscreen-keyboard users were fastest with prediction (Anson et al., 2006); with a simulated 1.5-second key delay, accurate prediction produced 59.9% more words per minute (Trnka et al., 2007) | Accessibility Keyboard with the Suggestions toolbar | Check compatibility before paying — WordQ lists onscreen keyboards and eye gaze as not supported |
| Typing is slow or painful with any tool | Across studies of people with physical impairments, speech recognition averaged 15.4 wpm, a standard keyboard 12.5, an onscreen keyboard with cursor selection 4.2 (with scanning, 1.7) (Koester & Arthanat, 2017) | Voice Control | — |
None of this is guaranteed for an individual. In Pouplin et al. (2016), 30 people with low-cervical spinal cord injury typed faster without prediction, whether the list showed 3, 6 or 8 words; the 15 with higher injuries saw no change either way.
And if a school or employer already licenses Read&Write, start there: it includes prediction on the Mac and now carries what Everway calls "Co:Writer word prediction" (Everway, 2026).
Which word prediction software runs on a Mac in 2026?
Three assistive-technology suites run natively on macOS (WordQ, Ghotit, Read&Write), two general autocomplete apps do (TypeTab, Cotypist), and Apple ships two free options. Co:Writer is retired, Mote runs only inside Chrome, and Lightkey is Windows-only. Prices checked October 2026:
| Tool | On a Mac | Predicts by sound | Where it works | Your text | Price | 3-year cost |
|---|---|---|---|---|---|---|
| macOS inline predictive text | Native, macOS 14+ | Not documented | Many apps; English, French, Spanish (current macOS) | Not stated by Apple | Free | $0 |
| Accessibility Keyboard suggestions | Native | Not documented | Onscreen keyboard | Not stated by Apple | Free | $0 |
| WordQ for macOS 6 | Native | Yes | Desktop apps, built-in PDF editor | "No data stored in the cloud"; online activation | CA$119/yr Standard, CA$149/yr Pro | CA$357 / CA$447 |
| Ghotit Real Writer & Reader | Native | Yes | Word, Pages, Safari, own editor | Perpetual licence works offline; subscription needs a network | $249/yr, or $399 once with 4 years of upgrades | $747 / $399 |
| Read&Write (Everway) | Native + Chrome | Not documented for Mac | Word, Google Docs, PDFs, Chrome, Edge | Desktop needs internet for most features | $184/yr individual | $552 |
| Co:Writer / Co:Writer Universal | Retired 31 Dec 2025 | — | — | — | No longer sold | — |
| Mote | Chrome extension only | Yes (vendor) | Chrome, Google Docs | Not stated | Free tier; $99/yr Individual | $0 or $297 — plan with prediction not stated |
| Lightkey | No — Windows 7–11 | — | Windows VM only | — | $49.95/yr or $229.95 lifetime | — |
| TypeTab | Native, macOS 14+, Apple Silicon | No | Almost all apps | On-device | Free trial, then $49 once | $49 |
| Cotypist | Native, Apple Silicon | Not documented | Every Mac app (vendor) | On-device | Free 100 words/day; $72/yr Plus; $108/yr Pro | $0 / $216 / $324 |
Sources: Apple Support, Apple Feature Availability, Quillsoft, Ghotit for Mac, Ghotit pricing, Everway pricing, Everway sunset notice, Mote extension, Mote text prediction, Lightkey FAQ, Lightkey pricing, Cotypist.
- macOS inline predictive text. Pick it if you want zero cost and type in English, French or Spanish — the languages Apple's current feature page lists (Apple, 2026). The catch: one short gray guess at a time, accepted with Space, and not every app shows it — the gaps are mapped in how to turn on predictive text on your Mac.
- Accessibility Keyboard suggestions. Pick it if you already type onscreen with a pointer, head tracking or Dwell. The catch: a list you click, not inline text, and users have reported the suggestions disappearing (fixes below).
- WordQ for macOS. Pick it if spelling is the barrier and you want phonetic prediction, read-aloud of suggestions and word banks in one app; the Pro dictionary targets post-secondary and workplace writing (Quillsoft, 2026). The catch: a yearly subscription, and onscreen keyboards, eye gaze and system speech recognition aren't supported. Version 6.3.8 shipped on 4 September 2026, so it's maintained.
- Ghotit Real Writer & Reader. Pick it if your misspellings are severe and sound-based and you'd rather pay once: its prediction is "designed for writers with typing issues and phonetic-based misspellings" (Ghotit). The catch: $399 up front, and the cheaper subscription is network-dependent (Ghotit, 2026).
- Read&Write. Pick it if someone else pays for it, or you also need read-aloud across Mac, Chrome and iPad. The catch: $184 a year for an individual, no individual trial on the pricing page, and phonetic prediction on the Mac isn't documented on the pages I checked (Everway, 2026).
- Mote. Pick it if the writing happens in Google Docs. The catch: it runs inside Chrome, so native apps like Mail, Pages and desktop Word are out of reach, and its pricing table doesn't say which plan includes Text Prediction (Mote, 2026).
- Lightkey. Skip it on a Mac unless you run Windows in a virtual machine — its own FAQ says it supports "MS Windows OS versions: 7, 8, 8.1, 10 and 11" (Lightkey).
- TypeTab. Pick it for fewer keystrokes in every app without a subscription. The catch: it doesn't predict by sound, and it needs Apple Silicon and macOS 14. Details further down.
- Cotypist. Pick it if 100 completed words a day covers you for free. The catch: unlimited use is $72–$108 a year, Apple Silicon only (Cotypist, 2026). Both autocomplete apps, and a few others, are compared in the best autocomplete apps for Mac.
How to turn on the free Mac options
All three are in System Settings and undo with the same switch. Inline predictions need macOS 14 Sonoma or later (Apple Support).
Inline predictive text (Apple Support)
- Open System Settings → Keyboard.
- Under Text Input, click Edit.
- Turn on "Show inline predictive text", then click Done.
Expected result: as you type in Notes, a gray completion appears after the cursor. Space accepts it; typing anything else ignores it. If nothing appears, check that your input source is an English, French or Spanish variant — those are the only ones Apple's current feature page lists (Apple Feature Availability) — and try another app, since Apple says "many", not all.
Onscreen keyboard on Mac: the Accessibility Keyboard with suggestions (Apple Support)
- Apple menu → System Settings → Accessibility → Keyboard → turn on Accessibility Keyboard.
- On the keyboard, click the Panel Options button (top right) → Toolbars → Suggestions.
- Panel Options → Typing → turn on automatic spaces and sentence capitalization.
- For eye or head tracking, Panel Options → Dwell.
Expected result: an onscreen keyboard with a row of suggested words; click one to insert it. If the suggestions stop appearing — one user with C4 quadriplegia reported exactly that on macOS Sequoia 15.5 — the fixes suggested in that Apple Community thread (2025) were: toggle the Suggestions toolbar off and on, delete and re-add the suggestions box in Panel Editor, and email accessibility@apple.com. The thread was never marked solved. To turn the keyboard off, click the close button in its top-left corner.
Voice Control (Apple Support)
- System Settings → Accessibility → Voice Control → turn it on.
- Click Open Tutorial to practise the commands.
Expected result: words you say in a text field are typed, because Dictation mode is on by default; say "Spelling mode" to dictate letter by letter. While Voice Control is on, standard macOS Dictation isn't available. Say "Stop listening" to pause, or switch it off in the same pane.
Does word prediction make you faster? Only when keystrokes are expensive
For fluent typists the evidence — mostly from phone keyboards — says it saves keystrokes and costs time. For people whose keystrokes are slow or effortful, often yes — but not reliably enough to skip testing.
| Study | Who typed, and how | What prediction did |
|---|---|---|
| Koester & Levine (1994) | 6 people with high-level spinal cord injury on their usual keyboard setup, 8 able-bodied mouthstick typists, 7 sessions | Slower for the spinal cord injury group, a modest gain for the others; keypresses slowed for everyone |
| Anson et al. (2006) | 10 adults on an onscreen keyboard | 7 fastest with word prediction, 2 with word completion, 1 no gain |
| Trnka et al. (2007) | 17 adults, onscreen keyboard with a 1.5-second delay per key | +59.9% wpm with accurate prediction, +10.1% with basic prediction |
| Koester (2014), in Koester & Arthanat (2017) | 9 single-switch users | 1.51 → 2.72 wpm after the setup was revised; no difference before |
| Pouplin et al. (2016) | 45 people with cervical spinal cord injury, 10-minute copy tasks | Low-cervical group faster without it; high-cervical group unchanged; list length didn't matter |
| Quinn & Zhai (2016) | Phone-keyboard transcription experiment | Fewer keyboard actions, preferred by users, slower overall |
| Palin et al. (2019) | 37,370 phone typists | Prediction use linked to slower typing (r = −0.183) and fewer keystrokes per character (r = −0.232) |
One mechanism fits most of the table. Each suggestion has to be noticed, read and judged — that's the cost. Koester and Levine measured it directly: keypresses got slower while prediction was on, and the spinal cord injury group searched the list much more slowly than the able-bodied group. The saved keystrokes are the payoff, and they're worth more the slower each keystroke is. Palin's team put the deciding factor in one phrase: "to which degree the user is rate-limited."
Prediction quality matters as much as the idea. With the same interface, Trnka's accurate engine saved users 50.3% of their keystrokes and the basic one 18.2%. A bad predictor is all overhead.
Keystrokes and speed are separate goals. If your hands are the limit, trading a few words per minute for far fewer keystrokes can be the right call. But none of the abstracts I could read report a pain result, and the one that asked about fatigue (Pouplin et al., 2016) reports no fatigue result in its abstract — so I can't tell you prediction prevents strain.
One more design question has partial evidence. List tools show several choices you scan; inline tools (Apple's inline predictions, TypeTab, Cotypist) show one guess at the cursor. I couldn't find a study comparing the two directly. Tam & Wells (2009) did find that moving the WordQ list down to keyboard level raised 21 young users' ratings of their own performance and satisfaction; they said the words were easier to find there.
Does word prediction software help with dyslexia and spelling?
In small studies of children, prediction programs with phonetic matching helped spelling or writing output for most participants. I couldn't find a controlled study in adults with dyslexia, so treat this as "worth a trial", not "proven".
The early tools failed exactly the writers they were bought for. Summarizing MacArthur's 1998 work, Evmenova et al. (2010) note that students with severe spelling problems "did not know the correct initial letters" — so the word they wanted never appeared in the list.
Evmenova's team then tested three phonetic-aware word prediction programs (Co:Writer, WordQ, WriteAssist) against plain word processing with six students in grades 3–6 over a month-long summer program:
- On average, 80% of prediction sessions beat every word-processing session on total words, and 84% beat them on composition rate (a single-case statistic called PND).
- Different students did best with different programs; four of six preferred WordQ.
- The design had no return to baseline, which the authors say limits the conclusions.
Handley-More et al. (2003) followed three children in grades 4–5 with learning disabilities and handwriting problems. Two clearly improved their spelling with word prediction; two still wrote fastest by hand. Their conclusion is the practical one: "evaluate each child individually."
Here's a 30-second test for any trial. Type three words the way they sound — nolij, sertin, fizishun — and look for knowledge, certain and physician in the suggestions. If they still don't appear once you've typed the whole sounded-out word, the tool probably matches spelling rather than sound, so it may not help with the words you get stuck on. WordQ and Ghotit also read suggestions aloud, which helps you pick between look-alike words (Quillsoft, 2026; Ghotit).
Test a paid tool for one week before you buy
Run the same 10-minute test without and with the tool, on day 1 and day 7, and keep it only if the number you care about moved. The method borrows from the studies above: Pouplin et al. (2016) used 10-minute copying tasks, Evmenova et al. (2010) used daily writing prompts.
- Pick your measure. Speed or effort: a copy task. Spelling: writing from a prompt.
- Copy task. Copy an old email or article for 10 minutes. Divide the word count by 10 for words per minute, count errors, and rate how your hands feel from 0 to 10.
- Prompt task. Write for 10 minutes on a fresh prompt ("what I did last weekend" works). Count total words and misspelled words.
- Run it twice on day 1 — tool off, then tool on — with a different passage or prompt of similar length each time.
- Repeat on day 7. Day-one numbers include learning the tool; the 1994 study also ran seven sessions per person.
- Decide. Keep it if misspellings fell or effort dropped without a big speed loss. Drop it if you're slower and no less tired.
Trial lengths, checked October 2026: WordQ 30 days free; Ghotit a 30-day money-back guarantee; Cotypist a 30-day Pro trial; TypeTab 100 word and 20 full-line completions. Read&Write's pricing page offers schools a demo and lists no individual trial.
If the tool counts keystrokes for you, use it. TypeTab shows "Saved N keystrokes today" in the menu bar; Cotypist points users to Settings → Statistics for accepted completions (Cotypist).
Where TypeTab fits — and where it doesn't
TypeTab is row two of the needs table: fewer keystrokes in every app, for someone who spells fine and whose hands are the limit. It shows your next words as faint ghost text in almost any text field — Mail, Messages, Slack, Notion, your browser. Tab takes the next word, Ctrl+E the whole line, and typing anything else makes the suggestion vanish. One suggestion at the cursor, nothing to scan.
It learns your vocabulary, names and phrasing on your Mac, works across many languages, and keeps every keystroke on-device — no account, nothing uploaded, and it works offline once the model has downloaded on first launch. The Stats pane tracks keystrokes saved, words completed and time saved, which makes the one-week test above easy to score. The trial is 100 word completions and 20 full-line completions with no card; after that it's $49 once, for up to three Macs.
Honest limits, since I'm the vendor:
- It doesn't predict by sound. It isn't built for phonetic spelling: it completes from the text you've already typed, and has no sound-based matching. If the first letters are the problem, WordQ or Ghotit is the right tool.
- It isn't sold as assistive technology, and I make no medical claim: fewer keystrokes is the whole pitch, and I found no study showing that it reduces pain.
- It has a hardware floor. Apple Silicon (M1 or newer), macOS 14 Sonoma or later, 16 GB of memory recommended. Intel Macs aren't supported, and there's no Windows, iPhone or iPad version.
Word prediction trades attention for keystrokes. Make that trade only where your keystrokes are the expensive part.
FAQ
What is word prediction software? Word prediction software suggests words as you type, based on the letters you've entered and the words before them; you pick a suggestion to insert it. Assistive-technology versions such as WordQ and Ghotit add phonetic matching, read-aloud and topic word banks. Mainstream predictive text software, like Apple's inline predictions or a system-wide autocomplete app, shows one inline guess you accept with a key.
Is there free word prediction software for Mac? Yes. macOS 14 and later include inline predictive text; Apple's current feature page lists it for English, French and Spanish (Apple Feature Availability). The Accessibility Keyboard has a Suggestions toolbar (Apple Support). Cotypist's free tier covers 100 completed words a day (Cotypist). Apple doesn't document sound-based prediction for its options; Mote advertises phonetic prediction, but its pricing doesn't say whether the free plan includes it.
What happened to Co:Writer? Everway — the new name for Texthelp, which acquired Don Johnston — retired Co:Writer and Snap&Read on 31 December 2025 and moved their features into Read&Write (Everway). An individual Read&Write licence costs $184 a year, checked October 2026 (Everway pricing).
Is WordQ available for Mac? Yes. WordQ for macOS 6.3.8 shipped on 4 September 2026, with orthographic and phonetic prediction, a 30-day free trial, and subscriptions from CA$119 a year (Quillsoft). Quillsoft says onscreen keyboards, eye gaze and system speech recognition aren't supported, though WordQ doesn't interfere with them.
Does word prediction help with dyslexia? In small studies of children it improved spelling or writing output for most participants (Evmenova et al., 2010; Handley-More et al., 2003). Results varied from child to child, and I couldn't find a controlled study in adults. Run a one-week trial with a spelling count before paying.
Does Lightkey work on a Mac? Not natively. Lightkey supports Windows 7 through 11, and its FAQ suggests running it inside a Windows virtual machine such as Parallels on a Mac (Lightkey).