The short answer
- On a phone, dictation is much faster than typing. In a Stanford-led study published in 2018, people entered English text at 153 words per minute by voice versus 52 on the iPhone keyboard: 2.9 times faster.
- Today’s best speech recognition gets about one word in twenty wrong on standard English benchmarks (5.63% for the top system in a 2025 comparison). Fine for a note; not fine if the wrong word is the amount.
- iPhone and Android both have built-in dictation: tap the microphone on the keyboard. On iPhone, dictation is processed on the device in many languages.
- For an expense, always say the amount, the currency, what it was and how you paid, then check the entry before saving it.
- On privacy: mute the mic when you don’t want to be heard, tell guests, review and delete your voice history, and be careful with children’s recordings.
Why voice works for small, everyday notes
Most expenses and ideas that never get written down have something in common: your hands were full, or you had two minutes. Pull out the phone, open an app, find the right field, thumb in “12.50”… That’s enough friction to think “I’ll do it tonight”. Tonight, you’ve forgotten.
Voice cuts that friction. In 2018, Sherry Ruan, Jacob Wobbrock, James Landay and colleagues (Stanford, the University of Washington and Baidu) compared typing and speaking short messages on an iPhone 6 Plus in a lab. In English, speech reached 153 words per minute against 52 on Apple’s keyboard, 2.93 times faster; in Mandarin, 123 against 43. Speech also made fewer errors during entry (5.30% vs 11.22% corrected error rate) but left slightly more in the final text (1.30% vs 0.79%). In other words: faster, but you still need to read it back.
Copying a given sentence isn’t the same as saying what you think. In 2020, Margaret Foley, Géry Casiez and Daniel Vogel repeated the comparison with 28 participants and added a task where people composed their own sentences. Speech was still faster, but the gap narrowed: 29% faster when composing versus 45% when transcribing. Thinking takes time, and dictation doesn’t remove it.
How accurate is dictation in 2026?
Researchers measure speech recognition with the word error rate (WER): the share of words substituted, dropped or added compared with what was said. In October 2025, the Open ASR Leaderboard, a public and reproducible benchmark, compared more than 60 systems on 11 datasets (first version of its reference paper).
- In English, the best system averaged 5.63%; OpenAI’s widely used Whisper Large v3 averaged 7.44%.
- In other major European languages (German, French, Italian, Spanish, Portuguese), top systems mostly landed between about 3% and 7%.
Two caveats. These numbers come from benchmark recordings (audiobooks, speeches, talks, meetings, earnings calls), not a noisy bus or a busy kitchen. And a 5% error rate means a twenty-word sentence contains one error on average. On an idea, that’s harmless. On “$45” heard as “$35”, it isn’t.
The dictation already on your phone
You don’t need a special app to start: the iPhone and Android keyboards can already take dictation.
On iPhone. Go to Settings > General > Keyboard and turn on Enable Dictation. Then tap the microphone key in any text field. According to Apple Support, dictation is processed on your iPhone in many languages, and no internet connection is required. In supported languages it adds commas, periods and question marks automatically; you can also say “new line” or “new paragraph”, or spell a word out letter by letter. It stops after 30 seconds of silence. One catch: when you dictate in a search field, the text may be sent to the search engine.
On Android (Gboard). Tap a text field, then the microphone at the top of the keyboard. When you see “Speak now”, talk. Google lists spoken commands such as “period”, “comma”, “question mark”, “new line” and “new paragraph”, and notes that punctuation isn’t available in every language. To fix a word, touch and hold it, tap the microphone and say the replacement.
Keyboard dictation types text, nothing more. To turn a sentence into a categorized expense or a reminder at the right time, you need a tool that understands the sentence, like an AI personal assistant.
Dictating an expense without mistakes
Most money-related dictation errors come from three places: the amount, the currency and the date. A four-part habit prevents them.
- 1The amount, spoken naturally“Twelve fifty” is ambiguous; “twelve dollars fifty” isn’t. For big numbers, say “three thousand five hundred” in one go.
- 2The currencyDollars, pounds, euros, rand, naira: say it every time, especially when you travel or keep accounts in more than one currency.
- 3What and how you paid“for a taxi, paid cash” or “groceries, on my card”. That’s what sets the category and the account.
- 4Check before you saveLook at the number first, then the date. One second now saves a correction three weeks later.
A few more habits:
- Say the date if it isn’t today. “Last night, twenty-two pounds on pizza” keeps today’s totals right.
- One expense per sentence. “Twelve dollars of bread and forty of gas” may be read as one $52 purchase, depending on the tool. Two sentences, two entries.
- For a long receipt, take a photo instead. Our guides on how to photograph a readable receipt and record cash purchases pair well with dictation.
Dictating notes and reminders
For an idea, speak as you would to a friend, then reread the first sentence: that’s what will help you find the note later. For a reminder, precision beats style: what, which day, what time, and whether it repeats. “Remind me on the first Monday of every month at 8am to pay rent” leaves little room for interpretation; “remind me about rent soon” leaves a lot. We compare reminder apps in a separate guide.
Privacy: what regulators recommend
Two regulators offer useful benchmarks. France’s data protection authority, the CNIL, published a white paper on voice assistants in September 2020 with practical advice for users. In 2017 it had already noted that most assistants send requests to remote servers, often keep a history (transcripts, recordings, metadata), and can wake up by mistake when they think they heard the wake word.
In the US, the Federal Trade Commission and the Department of Justice announced in May 2023 that Amazon would pay a $25 million civil penalty over charges that it kept children’s Alexa voice recordings for years and undermined parents’ deletion requests. The FTC’s advice to consumers: learn how to manage your child’s privacy, check whether your state gives you a right to deletion, and switch off features you don’t use.
- I mute the microphone, or switch the device off, when I don’t want to be heard
- I prefer a physical mute button and push-to-talk over an always-on wake word
- I tell guests that conversations may be recorded
- I opt out of my recordings being used “to improve the product” if I don’t want that
- I review my voice history regularly and delete anything confidential
- I never dictate a password, PIN or card number, especially in public
For children, the CNIL advises keeping assistants out of their own spaces (bedroom, playroom), supervising their use and regularly deleting the history. If your assistant is built on generative AI, also check whether your conversations are used to train the model; our AI personal assistant guide covers what to ask.
With Binome360: one sentence, one entry, your OK
In Binome360, you talk to your assistant the way you’d send a message, typed or by voice. It prepares the entry (an expense, a reminder or a fact to remember) and shows it to you: nothing is saved until you confirm. That’s your built-in moment for the read-back.
I paid 18 pounds for a taxi in cash, and remember that the driver who waits at the station is called Tom
Prepared: one expense and one fact in your memory. Save them?
Check the amount before you confirm. Nothing is saved without your OK.
Try Binome360 for freeThe same move works for a reminder (“remind me Friday at 5pm to pick up the dry cleaning”) or a life note (“remember that Priya has the spare key to Mom’s place”). You can also attach a file, such as a photo of the receipt.
Frequently asked questions
What is the best voice assistant for taking notes?
For plain text, the one already on your phone is enough: iPhone Dictation or Gboard voice typing on Android. Siri or Google Assistant add simple actions. To turn a sentence into a categorized expense or a recurring reminder, you need a tool that understands and files what you say, and lets you confirm first.
Does dictation work offline?
On iPhone, yes in many languages: Apple says dictation is processed on the device and no internet connection is required. On other devices and apps it depends on the service, so check your keyboard’s help pages.
Is my voice assistant always listening?
Smart speakers and wake-word assistants listen continuously for their wake word, and the CNIL notes they can trigger by mistake. A keyboard’s microphone key only listens after you tap it. If in doubt, prefer push-to-talk.
How do I add punctuation when dictating?
Say the mark’s name: “comma”, “period”, “question mark”, “new line”. On iPhone, punctuation is added automatically in supported languages; you can turn that off in Settings > General > Keyboard.
In short
Dictation is one of the fastest ways to stop losing track of expenses and ideas, with roughly one wrong word in twenty under good conditions. Say the amount, currency, item and payment method, read it back before saving, and keep your mic and voice history under control.
First step: turn on dictation on your phone, dictate your next expense as you leave the shop, and check the number before you save.
Sources
- Sherry Ruan, Jacob O. Wobbrock, Kenny Liou, Andrew Ng and James A. Landay, “Comparing Speech and Keyboard Text Entry for Short Messages in Two Languages on Touchscreen Phones”, Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, 1(4), 2018: doi.org/10.1145/3161187; preprint, “Speech Is 3x Faster than Typing for English and Mandarin Text Entry on Mobile Devices”, arXiv, 2016: arxiv.org/abs/1608.07323.
- Margaret Foley, Géry Casiez and Daniel Vogel, “Comparing Smartphone Speech Recognition and Touchscreen Typing for Composition and Transcription”, CHI 2020: doi.org/10.1145/3313831.3376861; full text: hal-02919659.
- Vaibhav Srivastav et al., “Open ASR Leaderboard: Towards Reproducible and Transparent Multilingual and Long-Form Speech Recognition Evaluation”, arXiv, version 1, October 8, 2025 (Tables 3 and 4): arxiv.org/abs/2510.06961v1.
- Apple, “Dictate text on iPhone”, iPhone User Guide (iOS 27): support.apple.com/guide/iphone/dictate-text-iph2c0651d2/ios.
- Google, “Type with your voice” (Gboard Help): support.google.com/gboard/answer/2781851.
- CNIL, “À votre écoute : la CNIL publie son livre blanc sur les assistants vocaux”, September 7, 2020: cnil.fr.
- CNIL, “Les conseils pour configurer et utiliser son assistant vocal”, September 3, 2020: cnil.fr.
- CNIL, “Le fonctionnement des assistants vocaux en 5 étapes”, December 5, 2017: cnil.fr.
- Federal Trade Commission, “FTC and DOJ Charge Amazon with Violating Children’s Privacy Law by Keeping Kids’ Alexa Voice Recordings Forever and Undermining Parents’ Deletion Requests”, May 31, 2023: ftc.gov.
- FTC Consumer Advice, “FTC says: Amazon didn’t protect Alexa users’ or children’s privacy”, May 31, 2023: consumer.ftc.gov.
Also available in Français.