How to Stop Dictation From Autocorrecting Technical Terms
You say "kubernetes" and your Mac types "cuban eatables." You say "PostgreSQL" and get "post gray sequel." If dictation keeps butchering your code, product names and jargon, the problem is not you. Here is how to make it stop.
Key takeaways
- The error happens inside the speech model, before autocorrect ever runs, so turning off autocorrect alone will not fix it.
- A custom dictionary is the real fix: it tells the app your exact spellings are correct.
- Apple Dictation has no true custom dictionary, only Text Replacements, which are limited.
- On-device tools with a custom dictionary and AI cleanup handle names, acronyms and code far more reliably.
Why dictation mangles technical terms
Speech recognition works by predicting the most likely words for the sounds it hears. Those predictions are weighted toward everyday language, because that is what most people dictate. When you say a rare term like an API name, a library, or a client's brand, the model has almost no reason to expect it, so it swaps in a common word that sounds close enough. That is why "React" becomes "reacts," "npm" becomes "N P M" or "enpin," and a colleague named "Sindre" turns into "cinder."
This is not a bug you can toggle off. It is baked into how a general-purpose model ranks words. The trick is to give the system a reason to expect your specific vocabulary. If your dictation also cuts out mid-sentence while you pause to think, that is a separate issue we cover in why Mac dictation stops after a few seconds.
Fix it step by step
Work through these in order. The first two are quick wins in the built-in Mac tools; the last one is the durable fix for anyone who dictates technical language every day.
Turn off fighting autocorrect
Open System Settings, then Keyboard, then Text Input, and edit the input source. Turn off "Correct spelling automatically" so typed corrections stop overwriting terms you already fixed by hand.
Add Text Replacements for your worst offenders
In Keyboard settings, add a Text Replacement for each term that breaks most often, mapping a trigger you can say clearly to the exact spelling. It is a workaround, not a real dictionary, but it patches the top ten words fast.
Check that Dictation is actually on
Confirm Dictation is enabled in System Settings under Keyboard. If it behaves strangely or refuses to type, run through our Mac dictation fix guide before blaming your microphone.
Switch to an app with a real custom dictionary
Add your names, acronyms and product terms to a dedicated dictionary in BlaBlaType. The on-device model then recognizes them directly, so "PostgreSQL" and "kubectl" come out right the first time, in any app.
Apple documents its built-in setup in the Mac dictation guide, and there is more on its on-device dictation approach if you want the background.
Where Apple Dictation falls short
Apple Dictation is free and convenient, but it has no true custom dictionary. Text Replacements are the closest option, and they force you to memorize triggers instead of just saying the word. There is also no AI cleanup layer that understands context, so a term the model gets wrong stays wrong. The comparison below shows the gap for anyone who lives in technical vocabulary.
| Capability | Apple Dictation | On-device app with dictionary |
|---|---|---|
| True custom dictionary | No | Yes |
| Learns names and acronyms | Text Replacements only | Yes |
| AI cleanup of filler and grammar | No | Yes |
| Works system-wide in any app | Yes | Yes |
| Audio stays on your Mac | Mixed | Yes |
If you regularly bounce between an editor, Slack and an AI chat, the ability to dictate into any app with your own vocabulary loaded is what makes voice typing usable for real work. Remember, most people speak around three to four times faster than they type, so the payoff is large once the terms come out clean.
Dictate your jargon without the mangling
Add your names, acronyms and code terms once. BlaBlaType keeps every word on your Mac and cleans up the rest. No card needed for the trial.
Download for macOSBuild a dictionary that actually sticks
A custom dictionary only helps if you feed it the right words. Run through this checklist once, and your dictation accuracy on technical language jumps for good. Keep the list somewhere you can add to it as new terms enter your work.
Custom dictionary checklist
- Add the ten terms dictation gets wrong most often, exactly as they should appear.
- Include proper nouns: teammate names, client brands and product names.
- Add acronyms with their intended casing, like API, SQL, gRPC and OAuth.
- List library and tool names you say aloud, such as PostgreSQL, kubectl and Tailwind.
- Use a custom AI prompt so the cleanup step preserves casing instead of lowercasing everything.
- Review your transcripts weekly and add any new term that slipped through.
- Keep everything on-device so your private terms and client names never leave your Mac.
Keep accuracy high over time
Technical vocabulary is a moving target. New frameworks ship, clients rename products, and your team hires people with names dictation has never heard. Treat your custom dictionary as a living list rather than a one-time setup, and add terms the moment you notice a mistake. Because BlaBlaType runs on-device, none of that vocabulary is uploaded, which matters when your terms include unreleased product names or client identities under NDA. If you want the wider picture of what tools handle this well, see our roundup of the best dictation software for Mac in 2026, or compare plans on our pricing page.
Frequently asked questions
Why does Mac dictation keep changing technical words?
Dictation is tuned for everyday language, so rare words like API names, libraries and product names are swapped for common words that sound similar. A custom dictionary tells the app those exact spellings are correct so it stops guessing.
Can I add my own words to Mac dictation?
Apple Dictation has no true custom dictionary, though you can add text replacements in Keyboard settings. A dedicated on-device app like BlaBlaType lets you add names and jargon directly, so they are transcribed correctly every time.
Does turning off autocorrect fix technical terms?
It helps with typed text, but dictation errors happen inside the speech model itself, before autocorrect runs. The reliable fix is a custom dictionary plus on-device AI cleanup that knows your specific terms.