Short answer: Mac dictation replaces names with common words because it picks whichever spelling is statistically likelier, and it has never seen your contact's name. Fix it by adding the person to Contacts, creating a Text Replacement entry in System Settings, and re-downloading the dictation language.
You dictate a sentence about a colleague and it comes back wrong. Aisha becomes "Asia." Rajan becomes "Rajun" or "region." Siobhan becomes something that looks nothing like a name at all. You fix it, dictate the next paragraph, and the exact same substitution happens again. Then again tomorrow.
This is the single most reported frustration with dictation on the Mac, and it is also the one people most often misdiagnose. It is not a microphone problem, it is not your accent, and in almost every case it is not a bug. It is a direct consequence of how dictation decides what you said. Once you understand the mechanism, the fixes stop feeling like superstition and start being predictable.
Why this happens: dictation is guessing, and names are the worst possible guess
Speech recognition does not hear letters. It hears a stream of sound, breaks it into candidate units, and then asks a much more important question: given everything that came before, which sequence of real words is most likely to be what this person meant?
That second half is why dictation feels so good on ordinary prose. If you say "I'll send you the deck before the meeting," the system does not need to nail every syllable, because only one common sentence fits that shape. Context rescues ambiguity constantly, and you never notice it happening.
Names break the mechanism in two ways at once.
First, a name carries almost no contextual constraint. "I talked to ______ yesterday" accepts literally any name in the world. The surrounding words give the system nothing to narrow with, so it falls back on raw acoustic similarity plus general frequency.
Second, general frequency is precisely the thing working against you. Your colleague's name might be the most common word in your working life, but across the language as a whole it is vanishingly rare compared to whatever ordinary word sounds like it. "Aisha" and "Asia" are acoustically near-identical. One appears in millions of sentences. The other appears in a few. The system picks the likely one, every single time, with total confidence.
So the behavior you are seeing is not the software failing at its job. It is the software doing exactly its job with information that does not include the people you work with. That distinction matters, because it tells you what actually fixes the problem: not better audio, but better information about your vocabulary.
Fix 1: Add the person to Contacts (the highest-leverage step)
macOS dictation draws on names in your Contacts app to bias its recognition toward people you actually know. This is the single fix that resolves the most cases, and it is the one most people skip because it feels unrelated.
- Open the Contacts app.
- Create or edit a card for the person whose name keeps getting mangled.
- Spell the first and last name exactly as you want it to appear in your text, including any accented characters.
- Save, then quit and reopen the app you were dictating into.
Two details make or break this. The card must be in an account your Mac actually syncs and indexes, so if you keep contacts in a work directory that only lives in a browser, the entry does not exist as far as macOS is concerned. And the spelling on the card is what you get back, so "Jon" on the card will never produce "John" in your document.
It is also worth confirming that dictation is permitted to use this data. In System Settings, look under Privacy & Security at the Contacts entry and make sure system services have not been blocked wholesale by an earlier permission prompt you dismissed.
Give it a little time. The bias from Contacts is not always instant, and a restart after adding a batch of names is a reasonable thing to do before you conclude it did not work.
What this fix will not do
Contacts only helps with people. It does nothing for product names, internal project code names, client company names, medical or legal terms, or the seventeen acronyms your team invented last quarter. Those need the next fix, and even then only partially.
Fix 2: Text Replacement for names Contacts cannot save
macOS has a built-in substitution table that fires as text is inserted. It was designed for shorthand expansion, but it works as a blunt correction layer for names that dictation reliably gets wrong in the same way every time.
- Open System Settings, then Keyboard, then Text Replacements.
- Click the plus button.
- In the Replace field, enter the wrong output dictation keeps producing.
- In the With field, enter the correct name.
The reason this works at all is that dictation is consistent in its errors. It does not roll dice. If it turned Kavya into "cavy a" this morning, it will do the same thing this afternoon, which means a single mapping catches it forever.
Now the honest caveats, because this fix has sharp edges.
It is a whole-string replacement, not a smart one. If the wrong output is a real word that you also use normally, you have just broken that word everywhere. Mapping "region" to "Rajan" will wreck every sentence about sales regions. Only use this when the erroneous output is something you would never legitimately type, which is more often the case than you would expect, because mangled names tend to come out as odd non-words.
It does not work everywhere. Text Replacement is a system text-input feature, and apps that draw their own text fields, including many browser-based editors and some Electron apps, ignore it entirely. If your corrections apply in Notes but not in your web CRM, this is why. Our post on dictation not working in third-party apps covers the same architectural split in more depth.
It also cuts the other way. An old text replacement can itself be the saboteur. If you once created a lowercase shorthand and dictation now produces that trigger string, you will see mysterious substitutions that have nothing to do with speech recognition. Scan the list for stale entries before adding new ones. This is the same mechanism behind the surprising results in our guide to Mac dictation not capitalizing.
Fix 3: Re-download the dictation language
If names started degrading suddenly, particularly after a macOS update, the on-device speech assets may be incomplete or corrupted. A partially downloaded model degrades gracefully rather than failing loudly, which is why the symptom shows up as "everything is slightly worse" instead of an error message.
- Open System Settings, then Keyboard, then Dictation.
- Note the language currently selected.
- Switch the language to something else, then switch back.
- Wait for the download indicator to complete before testing. This can take several minutes on a slow connection.
While you are in there, check the variant. English (United States) and English (United Kingdom) and English (India) carry meaningfully different name expectations. If you work with names common in one region while running the language variant of another, you have handed the system the wrong prior. Choosing the variant that matches the names you say most often is a bigger lever than most people realize, and it costs nothing to test.
Fix 4: Say the name inside a sentence, never alone
This is a habit change rather than a setting, and it produces a real improvement.
Dictating a bare name with pauses around it gives the recognizer the least possible context, exactly when it needs the most. Dictating the name inside a natural, continuous sentence gives it grammatical structure to work with, and structure narrows the candidate set. "Send it to Priya on Thursday" recognizes better than "Priya" said alone in a quiet room.
The same principle explains why slowing down and over-enunciating usually makes accuracy worse, not better. Isolated over-articulated words look unnatural to a system trained on connected speech. Speak at a normal conversational pace, in full phrases. We go deeper on this in Mac dictation inaccurate and typing wrong words.
Fix 5: Rule out the mic path before blaming the vocabulary
One class of name errors really is an audio problem, and it is worth ten seconds to exclude it.
If your Mac is routing input through a Bluetooth headset, the connection drops into a low-bandwidth call mode when the mic is active. Names are short and consonant-dense, which makes them the first casualty of reduced audio quality. Long common words survive the downgrade; "Beth" versus "Seth" does not.
Check System Settings, then Sound, then Input, and see what device is actually selected. Try dictating the same name with the built-in mic instead. If the built-in mic is dramatically better, your problem is the headset, and our guide to dictating with AirPods explains what to do about it.
The other audio check: run the test in a plain text field in Notes rather than in the app where you noticed the problem. If names are correct in Notes and wrong elsewhere, the recognition is fine and something in the destination app is interfering.
Fix 6: Correct by selecting, not by deleting
When dictation produces the wrong name, most people select the word and retype it. That is fine for the immediate sentence. But if you instead use the correction interface where available, choosing the right alternative rather than blowing the text away, you at least give the system a signal about the intended output.
Be realistic about the ceiling here. Stock macOS dictation does not maintain a durable personal vocabulary that grows from your corrections. That is the crux of the whole problem, and it is worth stating plainly rather than dressing up.
The part no setting fixes
Work through everything above and you will genuinely improve. Contacts catches the people. The language variant catches a whole class of regional mismatch. Text Replacement patches the stubborn repeat offenders one at a time.
But notice what you have actually built: a manual maintenance job. Every new client, every new hire, every product launch adds another entry you have to remember to create, in a table that does not apply in half your apps, keyed to an exact misspelling you have to notice and transcribe first.
And the underlying behavior never changes. The system still has no concept of your vocabulary. It has a list of contacts and a find-and-replace table bolted on afterward. The moment you say a term that is not in either, you are back to the statistically likely word, which is not the word you said.
Stock dictation does not learn. Every correction you make is spent on one sentence and then forgotten.
What a personal dictionary changes
The structural answer is dictation that knows your words before you speak them, and keeps knowing them.
Voice Keyboard Pro is a Mac menu bar app built around that idea. You hold a hotkey, speak, release, and the text appears at your cursor in whatever app you are in, system-wide, including the browser-based tools where Text Replacement quietly does nothing.
The relevant feature here is Smart Vocabulary, a personal dictionary with replacement rules. You add the names, client companies, product code names, and field-specific jargon you actually use, and Voice Keyboard Pro's transcription engine treats them as expected vocabulary rather than improbable outliers. That is a different mechanism from a system-wide find-and-replace: the terms are known going in, not patched afterward, so you are not dependent on the error being identical every time.
It also survives the thing that makes the manual approach collapse. One list, maintained once, applying everywhere you type. New client this week? Add the name once and stop thinking about it. Our overview of how custom vocabulary learns your words walks through how people actually build these lists in practice, which is usually a handful of terms at a time rather than one big session.
The same vocabulary follows you to the iPhone keyboard, where the mic button sits right on the keyboard and works in any iOS app. Voice Edit there lets you speak a change to fix text you already have, which is a faster repair than the two-finger cursor dance when a name does come out wrong.
A ten-minute plan
- Open Contacts and add complete, correctly spelled cards for the eight or ten people whose names you dictate most often.
- Check System Settings, then Keyboard, then Dictation, and set the language variant that matches the names you actually say.
- Scan your existing Text Replacements for stale entries that could be causing substitutions of their own.
- Add replacements for the two or three names that fail identically every time, but only where the wrong output is not a real word.
- Confirm your input device in Sound settings is what you think it is.
- Change your habit: full sentences, conversational pace, no isolated names.
Then watch for a week and count what still breaks. If the list is short and stable, the manual approach is holding and you are done. If you find yourself adding a new entry every few days, or the corrections evaporate the moment you move into a browser app, you have hit the design ceiling rather than a configuration mistake.
At that point the fix is not another setting. It is dictation that carries your vocabulary with it. Voice Keyboard Pro has a free tier with daily limits, and Pro is $4.99 a month or $34.99 a year if you decide the names should stop needing your attention.
Frequently asked questions
Why does dictation get common words right but names wrong?
Common words are constrained by grammar and context, so the system has strong signals about what fits. A name slot accepts anything, which removes the constraint and leaves only acoustic similarity and general frequency, where an ordinary word will always outrank a rare name.
Does adding someone to Contacts really work?
Yes, for people, and it is the most effective single step. It does not extend to products, projects, companies, or technical terms, and it depends on the contact being stored in an account your Mac indexes locally.
Will Text Replacement break my normal typing?
It can. The replacement fires on the trigger string regardless of context, so mapping a real English word to a name will corrupt legitimate uses of that word. Restrict it to mangled outputs that are not real words.
Is my accent the reason names come out wrong?
Usually not on its own. If ordinary sentences transcribe well and only names fail, the issue is vocabulary rather than pronunciation. Setting a language variant that matches the naming conventions you use most is the accent-adjacent fix that actually moves the needle.
Does dictation learn from my corrections over time?
Stock macOS dictation does not build a durable personal vocabulary from your edits. That is exactly why the same name fails repeatedly, and why a maintained personal dictionary is the durable answer.
Why do my corrections work in Notes but not in my browser?
Text Replacement is a system text-input feature. Apps that render their own text fields, which includes many web editors, bypass it. Insertion that happens at the cursor at the system level is not subject to the same limitation.