Mac Dictation Not Working in Zendesk? 9 Fixes That Work

Short answer: Zendesk has no dictation of its own, so failures come from macOS, the browser, or focus. The three most common causes are a call tool still holding the microphone, a reply composer that never took the cursor, and text that appears but saves empty because the editor missed the change.

Support agents write more words per day than almost any other desk job, so when dictation stops working in the ticket queue it is not a small annoyance. It is the difference between clearing a backlog and not.

The first thing worth knowing is that Zendesk does not have a dictation feature. There is no setting inside the agent workspace to turn speech to text on or off, which means no Zendesk preference can be the cause and no Zendesk preference can be the fix. Everything that can break sits in one of three layers: macOS dictation itself, the browser or desktop app the agent workspace is running in, or where the text cursor actually is when you start speaking. This guide works through those layers in the order that resolves the most cases fastest.

Note that the agent workspace changes regularly and the exact placement of controls varies by plan and version, so treat any interface description here as approximate and look for the equivalent control in yours.

The thirty-second test that tells you which layer is broken

Do this before touching a single setting. It saves ten minutes of guessing.

1. Open TextEdit or Notes, click into a blank document, and dictate one sentence.
2. Go back to Zendesk, click into the search box at the top of the agent workspace, and dictate one word.
3. Open a ticket, click into the reply composer so you can see a blinking cursor, and dictate one sentence.

Now read the result:

• Failed in TextEdit too. This is not a Zendesk problem at all. Dictation is off, the trigger key is wrong, or the microphone input is wrong. Go to fixes 1, 3, and 4, and our general Mac dictation fix list covers the rest.
• Worked in TextEdit, failed in the Zendesk search box. The problem is the browser layer or the microphone is held by something else. Go to fixes 2 and 5.
• Worked in the search box, failed in the reply composer. This is a focus or rich-text-editor problem, which is the most common Zendesk-specific case. Go to fixes 6 and 7.
• Text appeared in the composer but the reply saved or sent empty. Straight to fix 7. That one has a specific cause and a specific workaround.

1. Confirm dictation is actually on and you know the trigger

Open System Settings, go to Keyboard, and check that Dictation is switched on. It turns itself off more often than people expect, usually after a major macOS update, sometimes after a management profile is pushed to a work machine.

While you are there, look at the shortcut. The default on recent macOS versions involves the Globe key, and the Globe key has a separate dropdown of its own that decides what a single press does. If that dropdown is set to change input source or show emoji, your press is being consumed before dictation ever sees it. Set the dictation shortcut to something unambiguous and test it. Function-key remapping utilities are a common culprit on support team machines, because agents tend to accumulate them for macros and snippets, and they will intercept the key silently.

If the shortcut is the whole problem, the dedicated write-up on the Mac dictation shortcut not working goes through every variant.

2. Check the microphone your Mac is actually listening to

System Settings, Sound, Input. Speak at a normal volume and watch the level meter move. If it does not move, dictation is listening to a device that is not picking you up. On a support desk this is almost always a headset: it is connected, it is selected, but the boom is muted at the inline control or the headset has switched into a low-quality two-way mode for a call and back again.

One point that confuses people: Zendesk does not need microphone permission for macOS dictation to work. The browser asks for microphone access so that Zendesk Talk can make calls. Dictation is handled by the operating system before the browser is involved, so a denied page-level microphone prompt does not block it. Granting it will not fix dictation either.

3. Make sure the dictation language matches what you are speaking

If dictation starts, you speak clearly, and you get confident but completely wrong words, the dictation language is set to something other than the language you are speaking, or to a different regional variant. Support agents working across regions hit this constantly because they switch input sources during a shift and the dictation language follows.

Check the dictation language list in Keyboard settings, keep only the languages you actually use, and if you genuinely work in two, learn the switch rather than leaving it on automatic. Our post on dictation typing in the wrong language has the full picture.

4. Look for the other app holding the microphone

This is the top cause in a support team and the one people check last.

Zendesk Talk, a softphone, a screen recorder, or a meeting client can hold the input device open. Some of them keep holding it after the call has ended, because the window is closed but the process is still running in the background. The symptom is distinctive: dictation appears to start normally, the indicator shows, you speak, and nothing at all arrives. Not wrong words. No words.

Muting does not release the device. Mute means the other end cannot hear you, not that the microphone is free. The test is to fully quit every app that could be using audio, using Quit rather than closing the window, then dictate again in TextEdit. If it works, you have found it, and the ongoing fix is either to quit those apps between call blocks or to give dictation a different input device from the one your phone tool uses. A separate USB headset for calls and the built-in microphone for dictation removes the contention entirely.

The same pattern shows up on video calls, which we cover in dictation not working during video calls.

5. Rule the browser layer in or out

If dictation works in TextEdit but not anywhere in Zendesk, work down this list.

Test a second browser. Open the agent workspace in Safari if you normally use Chrome, or the other way round. If dictation works in one and not the other, you have isolated it to a browser rather than to Zendesk.

Check the browser profile. Chrome and Edge keep permissions and extensions per profile. Support agents frequently run a work profile alongside a personal one, and a setting you remember changing was changed in the other profile. Confirm which profile the window belongs to before concluding anything.

Check managed policy. On a company Mac, visit chrome://policy (or edge://policy) and look for audio capture restrictions. A managed policy overrides the settings UI, so the interface can show microphone access as allowed while the policy blocks it. This affects Zendesk Talk rather than macOS dictation, but it is worth knowing which of the two you are chasing before you file a ticket with your own IT team.

Disable extensions temporarily. Snippet expanders, grammar checkers, CRM sidebars, and productivity extensions all inject themselves into text fields, and a few of them fight with dictated input. Turn them off as a group, test, then re-enable in batches until the culprit appears. Our Chrome-specific guide has more on this.

6. Find out where the cursor really is

The agent workspace is dense. A single ticket view can contain the reply composer, an internal note tab, the requester and ticket fields in the sidebar, a tags field, CC and follower fields, a macro search, a side conversation panel, and one or more apps rendered in the sidebar. Dictation types into whichever of those has focus, and several of them look inactive while still holding the cursor.

Watch for these specifically:

• The macro search. Open the macro or apply-macro control, get distracted, and the search field keeps focus. Your dictated sentence filters macros instead of writing a reply.
• The tags field. It accepts text and turns spaces into tag boundaries, so a dictated sentence becomes a spray of nonsense tags. Undo carefully.
• The public reply versus internal note tabs. These are separate editors. It is entirely possible for text to be landing correctly in the one you are not looking at.
• Sidebar apps in iframes. Third-party apps in the sidebar re-render on their own schedule and can pull focus away mid-sentence. If your first four words land and the rest disappear, suspect this.
• Nothing focused at all. If you clicked a ticket in the list and then started speaking without clicking into a field, there is no text target. Depending on your account, bare keystrokes may also be bound to agent shortcuts, in which case dictated words can trigger actions rather than type. The habit that prevents all of this is: click the field, see the cursor blink, then speak.

7. The reply that looks right and saves empty

This is the Zendesk failure that wastes the most time, because it does not look like a dictation problem at all.

The reply composer is a rich-text editor, not a plain text box. Editors like this track their own content and tell the page when it has changed. Most of the time dictated text triggers that notification exactly like typed text. Occasionally it does not, and the result is text that is visibly on screen while the application still believes the field is empty. You submit, and the reply goes out blank or the draft is lost when you switch tickets.

The workaround is small and reliable: after dictating, type one character with the keyboard and delete it. That single real keystroke forces the editor to register a change, and everything you dictated is then included. If you would rather not think about it, dictate into a scratch field or a plain text editor and paste in, because pasting also fires the change notification.

If you see this consistently rather than occasionally, check whether a snippet or grammar extension is sitting between you and the editor, since those hook the same events. The identical problem in other web tools is covered in our Salesforce and HubSpot guides, where the cause and the fix are the same.

8. Things that look like bugs and are not

Four behaviours get reported as broken dictation every week and none of them are.

Dictation stops after a pause. macOS dictation ends a session after a stretch of silence. If you stop to read the customer's last message mid-reply, it may have finished without telling you and the second half of your sentence goes nowhere. Speak in complete chunks, or re-trigger after each pause.

Placeholders and dynamic content do not expand when spoken. Placeholder syntax and macro insertion rely on real keystrokes and menu selections. Saying the name of a placeholder out loud produces the words, not the value. Insert those with the keyboard or the macro control, then dictate the prose around them.

Mentions and links do not resolve. Speaking a colleague's name in an internal note writes the name, it does not create a mention. The picker needs typed input. Same for anything that triggers an autocomplete list.

The first word is clipped. If replies regularly start mid-word, you are speaking before the microphone is live. Pause for a beat after triggering. Bluetooth headsets make this worse because the device has to switch audio modes first, which is also why the built-in microphone is often the better choice for dictation. See dictation cutting off the first word.

9. Restart in the right order

When nothing above has worked, escalate in this order, testing after each step:

1. Quit the browser completely with Command-Q, not by closing the window. A browser with a background process alive keeps the same stale state that broke dictation.
2. Quit every call, recording, and meeting app the same way, for the reasons in fix 4.
3. Log out and back into macOS. This clears the stuck audio session cases without a full reboot.
4. Restart the Mac. Genuinely fixes stuck dictation and audio daemons, which is why support scripts start with it.
5. Check management policy. On a managed machine, a configuration profile or Screen Time restriction can disable dictation entirely, and the toggle will refuse to stay on. If dictation switches itself off every time you enable it, stop debugging and ask your IT team, because nothing you do locally will hold.

A faster option than fixing this every week

All of the above is about built-in macOS dictation, which is free and fine for short bursts but was not designed for someone writing two hundred replies a day. The recurring complaints from support agents are consistent: it stops at pauses, it cannot be taught product names, and it has no idea what your company calls its own features.

Voice Keyboard Pro approaches it differently. You hold a hotkey, speak, and release, and the text arrives at your cursor in whatever field you are in, including the Zendesk reply composer. Because you control when it starts and stops, there is no silence timeout to trip over, which matters when you are reading a ticket thread while composing.

The feature that changes support work specifically is Smart Vocabulary, a personal dictionary with replacement rules. Product names, SKUs, internal tool names, plan tiers, and the surnames of colleagues you mention in internal notes get transcribed correctly every time, instead of being fixed by hand in every third reply. Agents who handle a technical product usually find this is the whole game, because generic transcription gets ordinary English right and gets your product vocabulary wrong.

On privacy, the description is short enough to repeat to a security reviewer: audio is not retained and transcript content is not sent to or stored on our server. Our server records only operational pings, such as whether a device is active. Ticket text stays where you wrote it.

There is also an iPhone keyboard with a built-in mic button, which covers on-call and triage away from the desk, and two-way translation across 24 languages while dictating for teams answering in more than one language. Read what it produces before sending, as with any translation. Voice Keyboard Pro has a free tier with daily limits, and Pro is $4.99 a month or $34.99 a year.

Frequently asked questions

Does Zendesk have built-in dictation?

No. At the time of writing there is no speech-to-text feature in the agent workspace, which is why every fix here is a macOS or browser fix. It also means that if dictation works in other applications, Zendesk itself is rarely the thing that is broken.

Why does dictation work in Gmail but not in Zendesk?

Almost always focus or the rich-text editor. A Gmail compose window has one obvious text target, while a Zendesk ticket has a dozen fields competing for the cursor. Run the thirty-second test at the top to separate the two cases.

Do I need to give Zendesk microphone permission?

Not for dictation. The browser microphone prompt exists for calling through Zendesk Talk. macOS dictation captures audio at the system level, before the page is involved.

My reply sent blank after I dictated it. What happened?

The editor did not register the dictated text as a change. Type one character and delete it before sending, which forces the change to register. Fix 7 explains it in full.

Dictation stops halfway through a long reply. Why?

Built-in dictation ends after a period of silence, and reading the customer's message mid-reply is exactly that silence. Either speak in continuous chunks or use a push-to-talk style tool where you decide when the session ends.

It broke right after a call. Is that related?

Yes, and it is the single most common cause on a support desk. The call application is still holding the microphone even though the call has ended. Quit it fully and test again.

Does any of this apply to the Zendesk mobile app?

The causes are different on iOS, where the keyboard and per-app microphone permissions are involved instead. Our iPhone dictation fix list covers that side.

The short version

Run the thirty-second test first, because it eliminates two thirds of the possibilities in half a minute. If it fails everywhere, the problem is macOS. If it fails only in the composer, it is focus or the editor, and the type-one-character-and-delete trick resolves the empty-reply case immediately. If it broke today and you were on calls today, quit the call app before you change anything else.

Once it is working again, the longer-term question is whether built-in dictation is the right tool for a job that involves this much writing. Our guide to dictating in Zendesk covers the workflow side, including what to speak and what to keep on the keyboard.