Dictation for Journalists and Content Creators

Yes, journalists use dictation, and it has a long history in newsrooms. Reporters on deadline have phoned in their copy for decades, and today the same habit appears as voice typing into a laptop. The strongest case for dictation is the first draft: speech entry is much faster than typing in controlled tests. The strongest case against relying on it is accuracy. Names, quotes and numbers are where voice drafts go wrong, and the audio from a source can be a privacy question as well as a transcription one. If you write for a living, use dictation for the draft, check every fact against the recording, and decide where the audio is processed before you record anyone.

We found no recent survey that measures how many journalists or creators use dictation today, so the claims below rest on the history of newsroom practice and on measured speeds and accuracy, not on usage counts.

Reporters dictated stories by telephone for decades, and newsrooms built their routines around it

Before reporters filed from laptops, they phoned in. Reporters on deadline would telephone in and dictate their notes to an editor ¹, and the phrase "Get me rewrite!" comes from that routine. Dictation was also a standard office method well beyond newsrooms. Throughout most of the twentieth century, business people recorded voice letters, and a secretary typed them up from the recordings ².

Speed writing was a trained skill as well. In the UK, the NCTJ, the body that accredits most journalism courses, made shorthand optional in 2016, yet a speed of 100 words per minute remained a requirement for students on its news route who wanted the gold standard pass ³. Shorthand stopped being one of the mandatory skills modules in that change ⁴. The bar for human transcription is higher still in US court reporting: to earn the RPR, a court reporter must transcribe dictation at 180, 200 and 225 words per minute with 95 percent accuracy on each test ⁵.

So the habit is old. What has changed is the engine that turns speech into text, and whether it runs on your device or on a server. The rest of this guide covers how to use a modern engine for drafting, where its output needs checking, and where your recordings go.

Dictated drafts sound like speech, which suits scripts, podcasts and newsletters

Text you speak has the rhythm of talk, and that is an advantage for scripts, podcasts, newsletters and anything meant to be read aloud. A sentence that works in the ear is often a sentence that a reader would find too loose on the page, so plan to edit in both directions.

Speech also carries habits that writing does not. Speakers buy time with filler words. Clark and Fox Tree proposed that speakers use uh and um to announce a delay in speaking: a minor one they expect after uh, and a major one after um ⁶. A dictated draft keeps those patterns, so search the draft for filler words and cut most of them. Our answer on how to stop saying um when you dictate covers the spoken side in more detail, and the dictation commands guide lists how to insert and edit punctuation by voice.

Speak in short runs. Apple's guide for macOS Catalina 10.15 advised speaking for no more than 40 seconds at a time ⁷. Newer versions may behave differently, so test the length on your own version. Short runs also make it easier to correct a passage without losing the sentence before it.

A practical test for any dictated script is to read the draft aloud once. Where you stumble, the sentence is too long or too tangled for a listener, and where you find yourself inventing a clearer phrase, write that phrase down. Dictation gives you the raw material, and reading aloud shows you which parts will hold up.

Script length follows speaking rate, so size a video by words per minute

A 10-minute script is about 1,400 to 1,800 words if you read at the pace of an audiobook. Audiobooks are spoken at 140 to 180 words per minute ⁸, and 10 minutes at that pace comes to 1,400 to 1,800 words ⁹. The reading-rate meta-analysis that reports this range also uses 150 words per minute as its working speech rate ¹⁰, which sits inside the same band.

Those figures describe a steady, prepared read. Your own pace will differ, so measure it rather than trusting a table:

1. Copy a passage of about 300 words, the kind of prose you would actually deliver, into a document.
2. Read it aloud at the pace you would use on camera, and time it with a stopwatch.
3. Divide the word count by the minutes. That is your rate.
4. Multiply your rate by the length you need, and cut or add to match.

Treat the result as an upper bound. A video script also holds pauses, cutaways, on-screen text and reactions, and each of these takes time without adding words. Leave room for them. The question of how many words a 10-minute video script should be has the same answer in more detail, and the words per minute entry defines the measure.

Bloggers and newsletter writers dictate ideas on the move, then edit at a desk

Dictation is at its strongest when the goal is getting words down quickly, away from a keyboard. In a controlled study on an iPhone, English text entered by speech reached 153 words per minute, against 52 words per minute on the built-in keyboard, 2.93 times faster ^{11 12}. The same study measured Mandarin Chinese at 123 words per minute by speech against 43 on the Pinyin keyboard ¹³.

At these rates, a 1,000-word draft takes about 6.5 minutes spoken and 19.2 minutes typed ¹⁴. That is a saving of about 12.7 minutes per 1,000 words of entry ¹⁴, and the study's rates already include the time spent correcting errors as the text was entered. For the same text, speech takes about two-thirds less time than typing, before you edit the draft afterwards ¹⁵.

Those numbers come with limits. The study used short messages read aloud, in a quiet room, on a 2016 recognizer ¹². Newer engines are likely more accurate, but the study did not measure long documents or the time spent editing a dictated draft afterwards. Budget the editing time separately, and treat the saving as the time you save on drafting only.

The practical habits that help most are simple. Keep the microphone clear, because Apple advises that the Mac's microphone should not be obstructed by objects, such as clothing or your body, while you are speaking ¹⁶. Capture the idea in a sentence, then move on. Save the editing for a desk. A how to dictate a blog post follows the same sequence, and the guide to how to dictate emails uses the same drafting method for shorter messages.

Interview recordings are a transcription job, not dictation

Dictation is you speaking words that become text. An interview is someone else's speech, recorded, that has to be written down. That is transcription, and the two tasks need different tools and different checks. A dictation engine listens for your words in real time. A transcription tool works on a finished recording, and it can be run later, after the interview, in a quieter place.

For a recording you made on a Mac, Apple's Voice Memos app can recognize and transcribe the speech in audio recordings ¹⁷. Other tools do the same job, and the choice depends on where the audio goes and how accurate the output is for your speakers. Accuracy is not even across speakers. A 2017 study of YouTube's automatic captions found lower accuracy for women and for speakers from Scotland ¹⁸, so check the transcript of a woman speaker or a speaker from Scotland more closely, not less.

Keep the original audio for every interview you transcribe. A transcript is a draft of what was said, and the recording is the record. When you quote someone, go back to the audio at the exact timestamp. Our guide to interview transcription covers workflows in detail, and the verbatim transcription and clean verbatim entries explain the choice between keeping every filler word and removing them.

Quotes, names and numbers are where voice drafts must be checked word by word

Every quote, name and number in a voice draft needs a check against the audio or a primary document. Engines differ in how often they make errors. In a 2025 9to5Mac test of one 7-minute-31-second episode, Apple's speech API had a word error rate of 10.3 percent, against 1.5 percent for Whisper Large V3 Turbo ^{19 20}. Word error rate counts the substitutions, deletions and insertions needed to match a correct transcript, and the word error rate entry explains how it is calculated. One episode is a small test, so use these numbers to choose which engine to check more carefully, not to decide that one engine is safe.

A bigger risk is invented text. In a 2024 study of OpenAI's 2023 Whisper service, tested on recordings from speakers with and without aphasia, roughly 1 percent of audio transcriptions contained entire hallucinated phrases or sentences that did not exist in the underlying audio ²¹, and 38 percent of the hallucinations included explicit harms, such as perpetuating violence or implying false authority ²². Rates vary by model and by recording, so do not assume a rate you read in a study applies to your audio. What it means for you is simpler: a sentence in a transcript that you cannot find in the recording is a problem, whatever the engine's average.

The Society of Professional Journalists tells journalists to take responsibility for the accuracy of their work and to verify information before releasing it ²³. Its code also says that neither speed nor format excuses inaccuracy ²⁴. Those two lines describe the core problem with voice drafts: the speed is the benefit, and the check is the job.

A check that works for dictated copy has four steps:

1. Play back every quotation against the recording, and match the words exactly, including any repeated words.
2. Confirm every name spelling against a document, a press release, a company register or the person's own channel.
3. Recheck every number, date, unit and percentage. Numbers are easy to mishear, so do not trust the first version of any figure.
4. Mark anything you cannot verify, and leave it out of the published piece until you can.

Punctuation is a smaller issue, but it changes meaning. Automatic punctuation varies by app, and the dictation punctuation guide shows how to add marks by voice where the app does not do it for you.

Source protection affects where a journalist's audio should be processed

Where audio is processed decides who else may see it, which matters most when a source is speaking to you in confidence. Apple's Mac Help says you can check whether general text Dictation is processed on your device and not sent to Siri servers ²⁵. Other tools behave differently. Aqua Voice says that for users with Privacy Mode turned off, it may store transcript data on its servers to improve the product ²⁶. Apple links Siri and Dictation request history, including transcripts, to a random rotating identifier rather than your Apple Account, and says it may keep that history for up to two years, with a small reviewed subset kept longer ²⁷.

Vendor terms change, so read the current policy for each tool you use, and check what happens to recordings after the transcript is made. Our guides on whether voice-to-text is safe and human review of voice recordings explain the questions to ask. For a source's recording, prefer a tool that processes audio on your device, delete recordings you no longer need, and do not upload a confidential interview to a service you have not read about.

Ethics is part of this as well as technology. The SPJ code tells journalists to consider sources' motives before promising anonymity ²⁸. A promise of confidentiality is only as good as the systems that handle the recording, so the processing choice and the promise have to match.

Yap, which we make, is an on-device dictation app for Mac and Windows. Read its privacy terms on the product page before you use it for source material, the same as you would for any other tool. Our offline dictation for Mac page explains what runs without an internet connection.

This guide is general information about tools and practice, not legal advice. If a recording could expose a source to legal risk, take advice from your newsroom's legal team before you record or process it.

Dictation drafts fast, but the checks and the audio handling take their own time

Dictation for the draft, human checking for the facts, and deliberate choices about the audio: that combination covers most of the work. The related guides cover the other parts of the job. Dictation for work and study shows how the same method applies in other jobs. How much time dictation saves at work puts the speed numbers into a working week, and writing a book by dictation covers longer drafts. For the question of whether dictation is a common newsroom tool, see whether journalists use dictation. For the terms used here, the copytaker, voice writing and shorthand entries explain the older methods, and the on-device speech recognition entry explains where the processing happens.