How to Dictate Emails Faster Than You Type Them

Dictating email is faster than typing it when you already know what you want to say. In a controlled study, English text entered by speech reached 153 words per minute on an iPhone, against 52 on the built-in keyboard, 2.9 times faster ¹. The gain is largest on long replies. It shrinks, and can reverse, when you try to compose by speech from nothing. The method below is to plan each email in three short parts, speak it in one pass, and then check the details that dictation gets wrong. For the wider case for dictation at work, see dictation for work and study. For the time you can expect across a whole job, see how much time dictation saves at work.

Dictating email saves the most time on long replies, not on one-line answers

A one-line reply such as "Thursday works for me" takes seconds either way. The saving grows with length, because it is the gap between two rates multiplied by the number of words. At the study's rates, 1,000 words take about 19.2 minutes to type ² and about 6.5 minutes to speak ³, a difference of about 12.7 minutes ⁴.

The three bars use three different kinds of work, and the table shows why they should not be mixed up:

Read the first two rows with care. The 153 figure ¹ comes from short phrases that participants transcribed in a laboratory, and the authors say that results outside the laboratory still need study ⁸. The same authors say that upper-bound speech recognition performance has greatly improved compared with prior systems ⁹. Newer engines may therefore be faster to dictate with than the study's figures suggest, but this study did not measure that. Your own rate will depend on your voice, your room and your recognizer. Test it on a week of real replies.

The third row matters most for email. Composing by speech runs at a small share of normal speaking speed. Composition at 8 to 15 words per minute is roughly 5 to 12 percent of speaking speed ¹⁰. Planning is also the expensive part of composing. In a 1978 study, about two-thirds of the time spent composing a letter went on planning, whether the letter was written, dictated or spoken ¹¹. Dictation saves time on the typing, not on the thinking. So speak a plan you have already made, not the thoughts you are still forming.

Office workers spend a large share of the week reading and writing email

For the heaviest email users, email takes about a fifth of a standard working week ¹². In Microsoft's 2023 Work Trend Index, the most active 25 percent of Microsoft 365 users spent 8.8 hours a week on email ¹³. That is about 22 percent of a 40-hour week ¹². Two limits apply. This is the top quarter of users, not the average worker, and we could not open Microsoft's own report, so the figure is as the news outlet reports it. Ask how much of your own week goes to email before you decide how much dictation will help.

To see what dictation does to that time, take an example day of ten replies averaging 120 words each, 1,200 words in all. This is an illustration, not a measured average, and it assumes the replies are already composed.

Interruptions are the other cost. In a 2008 laboratory study of office tasks, including email replies, people who were interrupted worked faster and finished sooner, but the speed came at a price: more stress, higher frustration, time pressure and effort ¹⁷. Participants reported significantly higher stress after only 20 minutes of interrupted work ¹⁸. That study tested interruptions by phone call and instant message, not switching between voice and keyboard. Switching modes all day may carry a similar cost, and batching your replies into one session reduces how often you switch. The batching section below explains how.

Speak each email in three parts: the purpose, the details, the ask

The method works because it gives the dictated draft a shape before you start. The steps below follow the question on how to dictate a professional email:

1. Decide the purpose in one sentence before you press the microphone. For example: "Confirm the Thursday review moves to 3 pm."
2. Speak the purpose first, as the opening line of the message. A reader who stops after one line should know why you wrote.
3. Speak the details in the order the reader needs them: dates, names, figures and what is attached.
4. Speak the ask last, with a deadline if there is one. Keep the ask on its own line so the reader cannot miss it.
5. Speak the sign-off, or insert your usual closing from a saved snippet, which is what a dictation template is for.

Here is a short example of the result, with the three parts in order:

Subject: Thursday review moved to 3 pm

Hi Dana, I'm writing to confirm that the Thursday product review moves from 1 pm to 3 pm, in Room 4B. The agenda and the Q3 figures are attached, and the new budget numbers are on slides 6 and 7. Could you confirm by end of day Wednesday that you can attend?

The first sentence states the purpose. The middle carries the details. The last sentence is the ask. Dictation wrote the sentences, but the structure came from a plan made before you pressed the microphone.

Keep one purpose in each message. A message that asks for three things can get an answer to only one of them. If you need to raise several issues, send several short messages, or put each in its own paragraph with its own ask.

Spoken punctuation and formatting commands differ between apps and systems, so check the commands for your own setup before you rely on them. Our guide to dictation commands lists what to expect.

Names, numbers and addresses are where dictated email goes wrong, so check them before sending

In the same phone study, speech produced fewer errors during entry than the keyboard (a 5.30 percent corrected error rate against 11.22 percent), yet it left slightly more errors in the final text (1.30 percent against 0.79 percent) ¹⁹. Errors that survive are easy to miss, because a name spelled the way it sounds, a figure one digit out, or a word that sounds like the one you meant can look correct on a quick read.

Before you send, read the message on screen rather than trusting your memory of what you said, because the screen shows what the recognizer wrote. Scan first for capital letters and digits, since those are the items a reader will check. Then go through this list in order:

• Names of people, companies and products, including their spelling and capitals.
• Numbers and amounts, such as prices, quantities, invoice numbers and phone numbers.
• Dates and times, including the time zone when the reader is in another one.
• Email addresses and links, which dictation rarely gets right without a check.
• Words that sound alike, such as "to", "too" and "two", or "affect" and "effect".

The word error rate is the standard way recognition accuracy is measured, and it is worth knowing when you compare tools. It does not tell you which errors matter to a reader, so the list above is the more useful check.

Accents change accuracy, so test with your own voice

Accuracy depends on the speaker as well as the software. In a 2019 study that personalized recognition models, word error rate fell by 35 percent relative for the accented group, who were non-native English speakers with about an hour of speech each ²⁰. A 2004 study of recorded oral-history interviews found no evidence that accent affected recognition ²¹, but that was one archive and one recognizer. The results disagree, so do not assume either answer for your own speech. Dictate three real emails, count the errors you fix, and decide from that. Our question on dictation for non-native English speakers at work covers the practical steps, and speech recognition and accents covers the research.

Dictated email can read more conversationally, which usually helps

Speech is looser than writing. In a corpus analysis reported by Ishikawa (2015), lexical density, the share of content words in a text, was 33 percent in speech and 42 percent in writing ²². In practice, an email you speak from a plan can read more like an explanation to a colleague. That is usually what a work email should do. The cost is filler. Words such as "so", "basically" and repeated openings creep in when you speak, so read the draft once for them and cut them. The glossary entry on filler words explains how dictation tools handle them.

Every major mail app and both desktop systems take dictation, with different limits

Setup depends on the program you write in and the system it runs on. The table gives the one check that matters for each, and links to the full steps rather than repeating them here.

Our answer on the best voice to text app for email gives a longer comparison. For most people the built-in tools are enough. A dedicated app earns its place when you want one dictation tool across every program, or when you want on-device processing everywhere you write. Yap, which we make, is an on-device dictation app for Mac and Windows. Our blog post on choosing a dictation app explains the trade-offs, and the download page has the current platform details.

AI cleanup can turn a spoken ramble into a sendable email, at a privacy cost worth knowing

Some tools now tidy dictation automatically, removing filler and reordering sentences. That helps with a rambling first pass, but it changes three things. A rewrite can change what you meant, so compare it with your plan before you send. Cloud-based cleanup tools send the text to a cloud model, so the text leaves your device. And some keep the audio, so check each tool's retention policy. The contrast is clear in Apple's wording: on a Mac, general text dictation is processed on the device ²⁷, which is a different situation from sending a draft to a cloud rewrite. For the wider picture, see is voice to text safe and AI dictation cleanup.

Batching replies into one dictation session beats switching between voice and keyboard

Set aside one block of time for email. Read the inbox first and list the messages that need a reply, then order them by deadline. Dictate each reply in the three-part pattern, check the names, numbers and addresses, and send. Batching gives you one switch in and out of the task, where switching between voice and keyboard all day gives you many. The interruption study above is the nearest evidence we found, though it tested interruptions rather than switches between voice and keyboard ¹⁷.

Skill also improves with use. In a longitudinal study, users became more efficient at dictation, navigation and error correction as they gained experience ²⁸. A week of short sessions on real replies will show you more about your own speed than one test.

Long sessions can strain the voice, and researchers have studied this. A 2007 study of voice fatigue had four groups of speakers do a speech recognition task for 2 hours ²⁹. Take breaks during long sessions, and dictate the replies that need your words while typing short ones. For more on working at this pace, see writing a book by dictation and dictation for academic writing, which apply the same plan-then-speak pattern to longer documents.

Start with one email today. Speak its purpose, details and ask, check the names and numbers, and send it. Then do the same for the next ten. That is the test that will tell you whether dictation is faster for your inbox.