Wise Transfer Screenshot: Distinguish
Fee, Rate, and Recipient Amount
Wise is famous for transparency: you see the mid-market exchange rate, the exact fee, and the amount your recipient gets — every number laid out before you confirm. That same transparency, the one banks spend decades avoiding, is what makes a Wise transfer screenshot harder for traditional OCR than a Venmo payment that hides every cost behind one flat number. A Wise screenshot displays four distinct values — the amount you sent, the exchange rate, the Wise fee, and the amount the recipient receives — and OCR reads all four of them without knowing which one is the fee and which one lands in your recipient's account.
The Four Numbers on a Wise Screen — and Why Three Have No Label
A Wise transfer confirmation shows more information than any other payment app — and that's a deliberate design choice, not an accident. When you send money through Wise, the confirmation screen displays a three-line fee breakdown: the amount you're sending in your source currency, the mid-market exchange rate applied to the conversion, the Wise fee deducted, and the amount the recipient receives in their target currency. Four numbers. Three distinct meanings. One screen.
This layout exists because Wise monetizes differently from a bank. Banks mark up the exchange rate and call the transfer "free" — they hide the cost inside a worse rate. Wise uses the real mid-market rate — the one you can verify on Google or Reuters — and charges a small, visible fee on top. That single design decision has a visual consequence: a Wise screenshot contains a bank's entire hidden cost structure, printed in plain text, for anyone to read. The fee is on the screen because Wise chose to show it. Banks keep it invisible because they chose not to.
The numbers you'll see on a completed transfer confirmation break down like this:
- You send — the amount in your source currency, before any deductions (e.g., €1,000.00)
- Exchange rate — the mid-market rate applied, shown to 4-6 decimal places (e.g., 1.1863)
- Wise fee — the visible fee, typically 0.33%-2% depending on the currency pair and payment method (e.g., €5.74)
- Recipient gets — the final amount in the target currency (e.g., $1,180.12)
These four numbers sit in proximity on screen, usually stacked vertically, sometimes separated by horizontal rules or card-style containers depending on whether you're looking at the app, the web confirmation, or the PDF receipt. The visual layout changes. The four numbers — and their distinct meanings — stay the same. On the app, the rate might appear inline between the sent and received amounts. In the emailed confirmation, it might be in its own section labeled "Conversion details." The web confirmation places it differently still. Same data. Three different layouts.
There's also the multi-currency account view — a different screen entirely. Wise isn't just a transfer service; it's a multi-currency bank account with local account details in 22 currencies. If you hold a balance in euros and convert to dollars from within your account, the confirmation screen looks different from a one-off transfer: it may show a running balance before and after, the conversion timestamp, and a transaction reference. The four-number fee breakdown is still there — but it's embedded in a different visual context.
The Semantic Trap That Catches Every Traditional OCR
OCR can read all four numbers on a Wise screenshot. It just can't tell you which one is the fee. This is not an accuracy problem — it's a semantic identification problem. Reading the characters "€5.74" off a screen is the easy part. Knowing that "€5.74" is a fee, not part of the amount you sent, is the hard part — and traditional OCR has no mechanism for making that distinction.
Think about what a raw OCR output looks like on a Wise transfer confirmation. You get a block of text: "You send €1,000.00 Exchange rate 1.1863 Fee €5.74 Recipient gets $1,180.12." The machine spat out every visible character, including the labels — but as undifferentiated text. You still have to manually identify which number is the fee, which is the rate, and which is the amount your recipient receives. At that point you've done the recognition; what remains is a sorting exercise — and sorting four numbers into four columns is exactly what you were trying to avoid.
The irony runs deeper with Wise specifically. Venmo shows one number — the gross amount. PayPal might show a gross amount and a separate fee line — two numbers, two common meanings. A Venmo screenshot gives you one data point that you know is the amount. A Zelle screenshot gives you the amount, date, and sender — same one-currency world. A Wise screenshot gives you four numbers in potentially two different currencies with an exchange rate sandwiched between them — the most information-dense payment confirmation of any app. More transparency on screen = more numbers to sort = more rows where OCR gives you text but not meaning.
The problem isn't that Wise screenshots are hard to read. It's that they contain four semantically distinct values sitting next to each other, and a tool that reads by position rather than meaning can't distinguish "€5.74 — this is a fee charged to the sender" from "€1,000.00 — this is the amount being sent." Both are euro-denominated numbers on the same screen. Only meaning tells them apart.
Getting Four Fields Into Clean Columns — by Meaning, Not by Position
The extraction model that works on Wise screenshots is the opposite of the one that works on a structured form. A structured form — an invoice template, a tax return — puts fields in known positions. You can build a zone map: "Amount is always in the top-right box." Works until the form format changes. Wise screenshots never had a fixed zone to begin with — the app redesigns, the web confirmation uses different spacing, the email receipt adds its own formatting. Position-based extraction fails on day one because there is no fixed position.
Semantic extraction flips the logic. Instead of telling the tool where each number sits, you tell it what each number means. You define the four columns you want:
| Column Name | What the AI Looks For |
|---|---|
| Amount Sent | The principal amount in the source currency — the number before deductions, labeled "You send" or equivalent on the confirmation screen |
| Exchange Rate | The mid-market conversion rate applied to the transfer — a decimal value typically shown to 4+ places, labeled "Exchange rate" or inline with the conversion breakdown |
| Fee | The Wise fee deducted — a smaller amount in the source currency, labeled "Fee" or "Our fee" on the breakdown. The number is smaller than Amount Sent and sits below it in the visual hierarchy |
| Amount Received | The final amount in the target currency — the number labeled "Recipient gets" or "They'll receive," in a different currency code from Amount Sent |
The AI doesn't look at pixel coordinates. It reads the label next to each value — "You send €1,000.00," "Fee €5.74," "Recipient gets $1,180.12" — and matches the label's meaning to your column name. A fee is a fee whether it appears in line two or line three of the confirmation screen. An exchange rate is an exchange rate whether it's shown as "1.1863" or "1 USD = 0.8789 EUR." The rendering changes. The meaning doesn't.
Files are processed securely and not stored.
For clear, high-contrast text on a standard Wise confirmation screen, recognition accuracy reaches up to 99%. These are machine-rendered numbers on a digital display — the type of content vision AI models process most reliably. A five-second spot-check on the fee and exchange rate columns is enough for most records. And the principle holds: when a value genuinely isn't in the image — the transfer status was still "processing" when you screenshotted, so no confirmed rate was shown — you get an empty cell, not a fabricated number. No guess is safer than a wrong fee.
From One Screenshot to a Cross-Border Ledger
A single Wise transfer screenshot is a one-time record. Ten of them — across multiple months, multiple currency pairs, multiple clients — is a data entry project measured in hours. The volume grows naturally because Wise is inherently a cross-border tool: freelancers invoicing international clients every month, businesses paying overseas suppliers on recurring contracts, expats transferring money home on a regular schedule, accounting firms reconciling quarterly cross-border payments for multiple clients. A Venmo user might screenshot a dinner split once a week. A Wise user might generate a transfer confirmation every time an invoice gets paid — and every confirmation has four numbers needing extraction, not one.
The downstream pattern is batch-first by definition: upload all your Wise screenshots accumulated over the month, define the four column names once, and get back one table where each row captures all four values. The same column definitions work regardless of the currency pair — a €→$ transfer and a £→¥ transfer fill the same columns because the AI reads the labels, not the currency symbols. The output is one spreadsheet, not four individual lookups. For the full reconciliation picture — matching extracted Wise amounts against bank deposits, accounting for fees across platforms — the approach is the same batch-to-ledger workflow that applies to batch-reconciling payment screenshots from any app into a ledger.
On the tax side, Wise transfer confirmations serve as supporting records under the same principle as other digital payment receipts — they establish the amount, date, and counterparty of the transaction. Per Wise's own help documentation, you can download a PDF transfer confirmation or export statements in XLSX, CSV, MT940, CAMT.053, or QIF format. The exported data can import directly into QuickBooks, Xero, Zoho Books, or Wave. But not every transfer needs a full accounting export — sometimes you just need four numbers off a screenshot, right now, into the spreadsheet you're working in.
Frequently Asked Questions
Can OCR read the exchange rate and fee off a Wise screenshot?
It can read the characters — "Fee €5.74" and "1.1863" will both appear in the OCR output. What it can't do is label them. The output is a block of undifferentiated text containing all four numbers. You still have to manually sort "€5.74" into the "Fee" column and "€1,000.00" into the "Amount Sent" column — which means OCR solved the reading problem but left the semantic problem untouched. The human sorting step after OCR is the bottleneck, not the OCR accuracy.
Does the Wise screenshot layout change between the app, web, and PDF receipt?
Yes. The app confirmation shows the four numbers in a card-style layout with the rate inline. The web confirmation uses a slightly different visual hierarchy. The emailed PDF receipt places "Conversion details" in its own labeled section. The data is identical across all three — the four numbers don't change — but the layout does. A position-based extraction tool would need a separate template for each variant. Semantic extraction uses the same four column names across all three because it reads labels, not coordinates.
Is the multi-currency account balance screen different from a one-off transfer screenshot?
Different visually, same four-number fee breakdown conceptually. A conversion done from within your Wise account balance shows the source balance before the conversion, the rate applied, the fee, and the resulting balance in the target currency. The surrounding UI has more information (running balance, conversion timestamp, reference number) but the core four-number structure remains. If you're extracting both transfer confirmations and balance conversions in the same batch, the same four column definitions pull the right values from either screen type.
Wise made a design choice that no bank was willing to make: show the real exchange rate and the real fee, side by side, so you know exactly what a transfer costs. That choice created a screenshot with more information on it than any Venmo, Zelle, or PayPal confirmation — and a data entry problem that's harder precisely because the information is more complete. The four numbers on a Wise screenshot are four answers to four different questions. Semantic extraction—column names that mean something, AI that reads labels, not coordinates—is the only approach that gives you all four answers without asking you to sort them first. Drop in a Wise transfer screenshot and see the fee, rate, sent amount, and received amount land in four separate cells.