Short answer: Greenhouse has no dictation feature of its own, but system-wide voice typing works in every text box it has. Click into a scorecard field, hold your dictation hotkey, and speak your evidence. Keep clicking the ratings and typing the numbers, and dictate everything that is a sentence.
Thin scorecards are not a discipline problem. They are an input problem.
Every recruiting team has the same pattern in their data. The scorecards submitted within an hour of the interview are detailed and specific. The ones submitted two days later are three words long, and the three words are usually "good culture fit." Nobody set out to write a useless scorecard. What happened is that the interviewer walked out of a forty-five minute conversation carrying a head full of specific observations, went straight into the next meeting, and by the time they sat down to write, the specifics had decayed into a vague impression. Typing four paragraphs of structured evidence takes eight minutes. Speaking the same four paragraphs takes ninety seconds.
That gap is the whole argument for dictating in Greenhouse. Not that typing is unpleasant, but that the cost of writing determines whether the record gets written at all while it is still accurate.
Greenhouse does not have dictation, and that is fine
Let us be clear about the starting point. Greenhouse is a browser application. There is no microphone button in the scorecard. There is no voice input in the interview kit, no dictation in the candidate activity feed, no speech option in the email composer. If you are looking for a built-in feature or a marketplace integration to enable, there is not one to find.
You do not need one. Every field in Greenhouse is a standard web text input, and system-wide dictation on a Mac inserts text wherever the cursor happens to be. The application does not know or care where the characters came from. This turns out to be the ideal arrangement for recruiting teams specifically, because it means nothing gets installed into your ATS, no admin approval is required, no vendor gets connected to your candidate pipeline, and no browser extension needs a security review. You are simply typing, with your voice.
The same holds on iPhone. If you approve scorecards or reply to candidates from your phone between meetings, a keyboard with a built-in mic button works in the mobile browser and in any app on the device.
Setup: sixty seconds on Mac
- Open the candidate in Greenhouse and go to the scorecard or note field you want to fill.
- Click into the field first. This is the step people skip. Dictation inserts at the cursor, and if the cursor is nowhere, the text goes nowhere.
- Hold your dictation hotkey, speak the whole thought, release.
- Move to the next field and repeat.
One setup detail worth handling before your first real interview: check that your dictation hotkey does not collide with a browser shortcut or a Greenhouse keyboard shortcut. Web applications claim single-key and modifier shortcuts, and a hotkey that fires an unintended action mid-interview is a bad way to discover the conflict. Pick something unambiguous, test it once in a throwaway note field, and forget about it.
On iPhone, add the voice keyboard in Settings, enable it, then tap the globe key to switch to it inside Safari or your ATS app, and tap the mic button. The one requirement is Full Access, which is what lets a third-party keyboard use the microphone at all.
The five surfaces worth dictating
1. Scorecard attribute notes
This is the highest-value target in the entire product, and the most consistently neglected. Every attribute on the interview kit has a rating and a text box under it. The rating gets clicked. The text box gets skipped, because writing evidence for six attributes is a real writing task and nobody budgeted eight minutes for it.
Dictated, each one is fifteen seconds. Click into the box, hold, and say what you actually observed: what you asked, what they said, what it tells you. Six attributes, ninety seconds, and a scorecard that a hiring manager can act on instead of guess at.
The quality difference is not subtle. Typed attribute notes tend toward conclusions, because conclusions are short and evidence is long. "Strong systems thinking" is four words and typing it costs nothing. Spoken notes tend toward evidence, because speaking a specific example costs no more than speaking a vague label. Evidence is what makes a hiring decision defensible later, and voice biases you toward producing it.
2. The overall summary and recommendation
The summary is where the interviewer's actual thinking lives, and it deserves to be more than a restatement of the rating. Dictate it as a short spoken argument with a shape:
- The headline. Where you land, in one sentence.
- The strongest evidence for. The specific moment that most moved you toward yes.
- The strongest evidence against. The specific thing that gave you pause, stated as an observation and not as a verdict.
- What you could not assess. The area you ran out of time on, so the next interviewer knows what to cover.
- What would change your mind. The question that would resolve your uncertainty.
That last item is the one almost nobody writes and everybody wishes they had. It converts a scorecard from a verdict into a handoff, and it is the difference between a debrief where five people repeat each other and one where five people cover five different things.
3. Candidate notes on the profile
The running note on a candidate profile is a different genre from the scorecard. It is context: what a referrer said, what came up on a screening call, why a candidate went on hold, what their timeline constraints are, what they said about compensation expectations. This is the institutional memory that keeps a pipeline coherent across months and across recruiters, and it is exactly the kind of writing that dies when typing is the only input.
Dictating turns a note from a task into a reflex. You get off the phone, click into the note field, and talk for thirty seconds while it is fresh.
4. Outreach and candidate email
Sourcing outreach has a well-known economics problem: templates are fast and get ignored, and personalised notes work and take ten minutes each. Dictation collapses that trade-off, because the personalised part of a personalised message is two or three sentences of specific observation about that person's background, and two or three sentences is thirty seconds of talking.
The same applies to the messages nobody enjoys writing. Rejection emails after a final round deserve more than a template, and the reason they usually do not get more is that writing a considerate rejection is emotionally effortful on top of being slow. Removing the slow part helps more than you would expect.
5. Debrief and hiring manager updates
The written update to a hiring manager, the weekly pipeline summary, the note explaining why a role's funnel is stalling: these are narrative writing, which is the format voice handles best. Talk it out in one pass, then trim. Overshooting and cutting is faster than composing tight prose from a blank field.
The boundary: speak the sentences, type the settings
Voice is for prose. It is not for structured values, and pretending otherwise produces exactly the kind of error that is hard to catch on review because it looks plausible.
Keep typing or clicking:
- Scorecard ratings and the overall recommendation. These are buttons. Click them.
- Compensation numbers, offer figures, equity amounts, and anything else where a single misheard digit changes the meaning entirely.
- Requisition IDs, job codes, and internal reference numbers.
- Email addresses, LinkedIn URLs, and candidate phone numbers.
- Dates and scheduling details, which are easy to garble and expensive to get wrong.
- Tags, stage names, and dropdown values.
Dictate: attribute evidence, summaries, recommendations, candidate notes, outreach messages, rejection and offer emails, debrief writeups, intake meeting notes, and job description drafts.
The rule that keeps this simple: if the field would accept a sentence, speak it. If the field would reject a sentence, type it. Numbers in particular have their own set of habits worth learning if you dictate them often, which we covered in how to dictate numbers and dates correctly. In an ATS, the safer default is simply to type them.
The three-minute habit
Here is the workflow change that matters more than any feature.
Block three minutes at the end of every interview. Not later that afternoon. Not end of day. The three minutes immediately after the candidate leaves the call, before you open anything else.
In those three minutes, dictate straight into the scorecard: evidence for each attribute, then the summary, then submit. You are working from short-term memory that is still intact, including the exact phrasing the candidate used, which is the detail that vanishes first and matters most.
The scorecard written three minutes after the interview and the one written two days later are different documents about different candidates.
This habit is only sustainable because speaking is fast. Nobody keeps a three-minute post-interview ritual that actually takes nine minutes. Set the ritual to the length of the task the tool makes possible, and it survives a week of back-to-back interviews.
Keeping dictated hiring notes defensible
Recruiting records are not casual documents. Interview notes and scorecards can be requested in employment disputes and are subject to your organisation's own record-keeping and data-protection obligations. Dictation does not change any of that, but it does change one thing about your behaviour, and it is worth naming.
Speaking is more casual than typing. It is looser, more conversational, and it produces asides that you would never have bothered to type. That is usually a benefit. In a hiring record it is a risk, because the aside that felt like harmless colour in the moment reads very differently in a printout two years later.
Three practices keep this clean:
- Talk about the job, not the person. Describe what the candidate demonstrated against the criteria on the interview kit. Personal characteristics, appearance, accent, family situation, age, and health have no place in a hiring record, and voice makes them easier to blurt out than typing does.
- Say observations, not diagnoses. "Could not explain the trade-offs in their own architecture when asked twice" is an observation. "Not smart enough for this team" is a diagnosis, and it is the kind of sentence that a dictation habit produces and a typing habit filters out.
- Read it before you submit. Every dictated note gets a ten-second read-through. This is good practice for accuracy anyway, and it doubles as the moment you catch the sentence you should not have said.
Your legal and HR teams own the actual policy on what may be recorded, retained, and where. That is their call, not ours. If you are introducing a new tool into a workflow that touches candidate data, ask them first. The same goes for recording interviews at all, which is a consent question with different answers in different jurisdictions and is entirely separate from dictating your own notes afterward. For a broader look at how voice input fits into people-team workflows, we wrote about it in voice dictation for HR professionals.
The names problem, and how to solve it once
Recruiting has a vocabulary problem that most professions do not. You dictate proper nouns constantly and they are the hardest class of word for any transcription system to get right, because a rare name always loses to a common word that sounds similar. Candidate names from every linguistic background. Company names. Framework and tool names. Internal team names and project code names that exist nowhere outside your company.
Speaking more clearly does not fix this. The engine is not mishearing you, it is picking the more probable word, and the more probable word is wrong. The only real fix is telling the system what your vocabulary is.
Smart Vocabulary in Voice Keyboard Pro is a personal dictionary with replacement rules for exactly this. You load in the names you say constantly, and they come out right from then on, in the scorecard and the email and the note, without you correcting anything. For a recruiter running one requisition for three months, that is the same twenty proper nouns hundreds of times, and fixing them once at the source is a different experience from fixing them one at a time forever.
Where Voice Keyboard Pro fits
Voice Keyboard Pro on Mac lives in the menu bar. Hold a hotkey, speak, release, and the text appears at your cursor in whatever window is in front of you. Because it inserts at the cursor system-wide, it works in Greenhouse the same way it works in your email client, your calendar, and your notes app. Nothing is installed into the ATS and nothing needs your IT team's approval to try.
Meeting Mode is the piece that fits the debrief half of the job. It listens to a meeting, distinguishes between speakers, and produces AI notes, without a bot joining the call and announcing itself to everyone in the room. For an intake meeting with a hiring manager, where the entire point is capturing what "good" means for this role so you can write the interview kit against it, that is the difference between arriving with notes and arriving with a memory. Calendar meeting detection means it can recognise when you are in a scheduled meeting rather than requiring you to remember to start anything. As with any recording or note capture in a hiring context, follow your organisation's policy and tell people what you are doing.
On iPhone, the app is a keyboard with a built-in mic button, so it works in the Greenhouse mobile experience and anywhere else you triage from your phone. Two features earn their place in recruiting specifically. Voice Edit lets you speak a change to text you have already written, which is how you fix a sentence in a rejection email on a phone screen without fighting the cursor. Two-way translation while dictating covers 24 languages, which matters if you are sourcing internationally and want to send a candidate a note in their language.
On privacy, and this is a fair question when candidate data is involved: our server stores only operational pings. No audio, no transcript content. What you dictate into a scorecard is not something we hold.
There is a free tier with daily limits if you want to run it through a week of real interviews first. Pro is $4.99 a month or $34.99 a year. If you want the wider picture of how voice input changes a recruiting workflow beyond one tool, we covered that in dictation for recruiters.
Frequently asked questions
Does Greenhouse have a built-in dictation or voice-to-text feature?
No. Greenhouse is a browser application with standard text fields and no microphone input of its own. System-wide dictation is what makes voice work there, which is also why it works identically in every other field the product has.
Do I need my IT team to approve anything?
Nothing is installed into your Greenhouse instance and no integration is connected to it, because a system-wide dictation tool is typing on your behalf at the operating system level. That said, introducing any new tool into a workflow that handles candidate personal data is a conversation worth having with your security and legal teams before you standardise on it across a team.
Will this work in the Greenhouse mobile app or mobile browser?
On iPhone, a third-party keyboard with a mic button works in any app that accepts keyboard input, including mobile browsers. You switch to it with the globe key and tap the mic.
Can I dictate the ratings as well as the notes?
You can say the words, but you should not. Ratings are buttons and radio selections, not text. Click them. Voice is for the evidence underneath.
What about interview notes taken live during the interview?
Dictating while a candidate is talking is a bad idea for obvious reasons. Type sparse keywords live, then dictate the full scorecard in the three minutes after the call, using your keywords as prompts. This is a better process than live note-taking anyway, because it keeps your attention on the candidate rather than on your keyboard.
Is dictated text lower quality than typed text?
Different, not lower. Dictated writing is more conversational and less compressed, which is a liability in a formal document and an asset in a scorecard, where specific and slightly rambling beats terse and vague every time. A ten-second read-through before you submit closes most of the gap.