How to Log Expenses by Voice on iPhone (One Sentence, Logged)
Quick answer
To log an expense by voice on iPhone, open a voice expense tracker, tap the mic, and say three things: the amount, what it was, and where. “Twelve dollars, lunch, at the deli” is enough. An AI model turns the sentence into a transaction with an amount, a merchant or note, a tag matched to your own list, and a date. You can cover a whole day in one recording, in any major language. Review the result, fix anything misheard with Edit, and fall back to the keypad when you have no connection.
The hard part of expense tracking was never the arithmetic. It is the thirty seconds after paying, when you weigh up whether logging this is worth opening an app for. A form loses that argument most days. A sentence wins it.
So the whole task becomes "twelve dollars, lunch, at the deli", and the tag, the date and the formatting are the AI's problem.
This page covers what to say, how words become a transaction, how to check the result, which languages work, what breaks without a connection, and the times the keypad is simply the better tool.
Why voice beats the keypad for everyday spending
A keypad entry is quick, three taps in a decent app, but it still asks you to switch into data-entry mode. Find the tag. Decide on a note. Confirm.
Speaking skips the mode switch entirely, because when you describe a purchase the way you would describe it to somebody walking beside you, that description already contains the amount, the place, the rough category and the day, which happens to be everything a tracker actually wants.
There is a quieter benefit too. Work on payment form, including Raghubir and Srivastava's research on payment coupling, found that spending feels less real the further it drifts from handing over cash. Saying an amount out loud, even to a phone in your pocket, puts a fragment of that back. Try saying "forty dollars on snacks" three days running. You notice.
Voice logging in Cash Book is the coral mic circle at the bottom of Home. It is one of four ways in, next to the camera, an Apple Pay automation and the keypad, and the whole system is laid out in how to track expenses on iPhone.
What to say: amount, what, where
Three pieces of information make a complete entry and the order genuinely does not matter. Say the amount however it comes out, "twelve fifty", "nine dollars", "four euros twenty", and skip the currency if it is the one already set in Settings. Say what it was, because "lunch" or "a haircut" or "the electricity bill" becomes the note and is the strongest clue for the tag. Then say where: at the deli, at the market, on the bus.
A few extras get picked up if you include them. Time words move the date, so "yesterday" and "on Saturday" work and anything without a time word lands on today. Naming an account, "on the cash", "from savings", puts it there instead of the default. And direction is inferred from how you phrase it, so "Ana paid me back twenty" comes back as income rather than an expense, which is the one that surprises people the first time.
The tag you never have to say.
Logging the whole day in one go
The end-of-day run-through is the habit worth building. Hold the mic on the walk home and say the day out loud in one breath:
"Coffee three fifty this morning, lunch fourteen at the noodle bar, twenty on the bus card, eight at the pharmacy, and I got the ten back from Sam."
Five transactions. Four expenses and an income, each with its own amount, note, tag and date, from one recording of about fifteen seconds. When several land at once the toast tells you the count and the total rather than listing them, so read the day's list underneath to check the split.
Mention the account when it matters, and the pharmacy can go on cash while the lunch goes on the card. How to track cash spending has more on that side.
How the AI maps words to tags
Every recording goes to an AI model through the app's relay with a little context attached, and the context is what makes the result feel like yours: today's date and weekday, your time zone, your currency, your tag names, your hashtags, your accounts and which is the default. Back comes a list of transactions, each with an amount, a currency, a merchant, a tag, a note, a date, and whether it is expense, income or transfer.
Matching works on meaning, not keywords, which is why "the gym fee" finds a fitness or health tag if you have one, "Lidl forty-two" finds groceries because Lidl is a grocer rather than because the word groceries was anywhere in the sentence, and "drinks with Tom" lands on eating out or fun depending entirely on which of those two your list happens to contain. It picks from your tags rather than inventing new ones, which is exactly why a short list with plain names outperforms a long list with clever ones, and why five minutes spent in tags and categories pays for itself here more than anywhere else in the app.
Over a few weeks your phrasing quietly adapts to what lands cleanly. Nobody plans this. It just happens.
Six example phrases and what they become
| You say | What lands in the app |
|---|---|
| "Twelve dollars, lunch, at the deli" | Expense, 12.00, note "Deli", tag Eating out, today |
| "Groceries forty-two at the market yesterday" | Expense, 42.00, note "Market", tag Groceries, yesterday's date |
| "Coffee three fifty and a croissant two" | Two expenses, 3.50 and 2.00, both tagged Coffee or Eating out, today |
| "Twenty on gas, from the cash" | Expense, 20.00, tag Transport or Fuel, on the cash account |
| "Sam paid me back the ten" | Income, 10.00, note "Sam", tag Paid back or your closest income tag |
| "Electricity bill eighty-six on Monday" | Expense, 86.00, tag Bills or Home, dated the most recent Monday |
Tag names depend on your own list. The point of the table is that none of these sentences have structure.
Review and edit before it sticks
Nothing gets written silently. The screen says "Reading your transaction…" for a moment, then the entries appear in the day's list on Home with Edit next to them. Read them once. Usually that is a glance and you move on. When one is wrong, Edit opens the same sheet the quick keypad uses, with the amount, tag chips, note, date, account and the Expense, Income or Transfer pill.
Two errors are worth watching for. Numbers that sound alike, fifteen and fifty being the eternal pair. And amounts that quietly went to the wrong account.
Merchant names take the worst beating. Say a small local place quickly, outdoors, into a phone at arm's length, and what comes back can be a phonetic spelling that belongs to no shop on earth. Amounts survive far better than names, which is the right way round.
Sometimes it fails outright: "Couldn't read that", and underneath, "We couldn't extract a transaction. Try again or add it manually." In my experience that is almost always a noisy room or a sentence that trailed off. Say it again, closer, shorter.
Languages and currencies
Voice understands spoken expenses in any major language, and mixing them mid-sentence is fine. A Spanish speaker naming a German supermarket gets the same result as an English speaker naming an American one. Amounts are read in the currency you say, so "forty euros" stays euros even when your app currency is dollars.
The interface ships in 26 languages, and Settings offers 128 currencies. In practice that means your tags, your screens and your spoken sentences all sit in one language, which is most of why tag matching feels effortless rather than approximate.
What needs a connection, and when to use the keypad
Voice needs a connection. The parsing happens through the relay to the AI model, not on the phone, and the same goes for snapping a receipt. No signal, no transaction.
The keypad is the offline path and always one tap away: black plus, a big number, tag chips in the footer, optional note, date and account. Three taps, anywhere, including a plane. It is also the better choice in a quiet carriage or an open-plan office, where saying "sixty for the physio" out loud is a slightly bigger event than you want it to be.
For anything already on paper, the camera wins. How to scan receipts into an expense tracker covers that.
Privacy: what happens to the recording
The clip goes through the studio's relay to an AI model, gets parsed, and is dropped. The transactions themselves live on your phone, plus your own encrypted iCloud if you turn sync on. Analytics count that a voice log happened and whether it succeeded, never the words and never the amounts. There is no sign-up and no account, so there is nothing to log in to and nothing much to leak.
Making it a habit
Speak it on the way out, while the amount is still on the card reader in front of you. It genuinely takes less time than putting the phone back in your pocket.
Keep one fixed catch-up moment for whatever got away, the walk home or the kettle or brushing your teeth, and run the day in one recording. And read one number a week: tap 1W on Home, look at the total, then once a month open Insights to see which tag moved. Voice entries are only worth making if something downstream reads them.
Small amounts are the ones people skip, and they are the ones that quietly accumulate. The small purchases calculator turns a daily "it's only three dollars" into an annual figure, which is worth seeing before you decide coffees are beneath tracking. Spoken, a coffee costs about four seconds to record and closes the gap that keeps most impulse purchases invisible.
What this is based on
Logging promptly and reviewing weekly follows public money-management guidance from MoneyHelper and the Consumer Financial Protection Bureau, both of which start budgeting from a complete record of what was actually spent. The Federal Reserve's Economic Well-Being of U.S. Households report describes how routinely households are surprised by their own spending. The link between how tangible a payment feels and how much goes out is from Raghubir and Srivastava's peer-reviewed work on payment form, listed below.
Get Cash Book on the App StoreFree for 7 days. $19.99 once for lifetime.
Frequently asked questions
What do I actually say to log an expense by voice?
The amount, what it was, and where, in whatever order comes out. “Nine fifty for a sandwich at Pret” and “Pret, sandwich, nine fifty” land the same. Add “yesterday” or “on Monday” to move the date. You never need to name the tag, because the AI is handed your own tag list and picks from it rather than inventing something new.
Can I log several purchases in one recording?
Yes, and this is the feature worth learning first. Run through the day the way you would tell a friend: “coffee three, lunch twelve at the noodle bar, twenty on the bus card and eight at the pharmacy.” Each item becomes its own transaction with its own amount, note and tag. It is by far the cheapest way to rescue a day you forgot to log.
Does voice logging work in my language?
It works in any major language, and mixing them mid-sentence does not break it: a Spanish speaker naming a German supermarket gets the same result as an English speaker naming an American one. Amounts are read in the currency you actually say. The interface ships in 26 languages, so the tags you speak to can match the language you think in.
How does it know which tag to use?
It matches on meaning against your list. “Groceries at the market” finds your groceries tag even if you named it Food shop; “the gym fee” finds fitness or health if such a tag exists. Nothing fits? It falls back to a sensible default and you change it with one tap. Short, plainly named tags work noticeably better than clever ones.
Is my voice recording kept anywhere?
No. The clip travels through the app's relay to an AI model for parsing, and nothing is stored once the transactions come back. Usage analytics count that a voice log happened and whether it worked, never the words or the amounts. There is no account to hold a history against anyway, because the app has no sign-up at all.
What happens if I have no signal?
Voice needs a connection, since the parsing happens through the relay rather than on the phone. With no signal you get an error instead of a transaction. Use the keypad: black plus, type the amount, tap a tag, done, three taps and no connection required. Afterwards nobody can tell which entries were spoken and which were typed.
What if it hears the wrong amount?
Every voice log lands visibly, with an Edit option sitting right there. Open the entry, fix the number, tag, note, date or account, save. Nothing is written behind your back. The classic mishear is “fifteen” against “fifty”, which is a two-second correction today and a genuine mystery three weeks from now.
What this is based on
Get Cash Book on the App StoreFree for 7 days. $19.99 once for lifetime.