How to Extract Korean Receipt
Data to Excel for Tax Filing
The NTS already has digital records of every cash receipt (현금영수증) your business received last year. The card companies have electronic logs of every swipe. Between the two, the government knows about most of your deductible expenses before you do. And yet every May, the bottleneck in filing comprehensive income tax (종합소득세) as a freelancer or small business owner is the same as it was a decade ago: turning three different receipt types — digital, printed, and handwritten — into a single categorized spreadsheet that your tax accountant (세무사) can actually use.
Key Takeaways
- The NTS already holds digital records of every cash receipt and card swipe your business generated last year — but you still spend 18 to 30 hours every May retyping the same data into a spreadsheet.
- Half of those hours are not data entry — they are the second pass where you go back through 500 rows and manually label each receipt as Meals, Transport, or Office Supplies.
- ImageToTable.ai extracts date, vendor, amount, and category in a single pass — you trade 500 manual entries and a second categorization pass for a spreadsheet that arrives pre-sorted and ready for your 세무사.
The Three Receipt Types Korean Freelancers Actually Deal With
Before you can extract anything, you need to know what you are working with. Korean business expenses generate three distinct receipt formats, and each one creates a different data problem.
Cash receipts (현금영수증) are the best case. Issued through the NTS's dedicated 현금영수증 system, every cash transaction above a trivial amount generates a digital record automatically transmitted to the tax authority. You can view them on HomeTax (홈택스) and download a CSV file containing transaction date, amount, vendor business registration number, and approval code. The data exists — but the export format is aggregate, not item-level. A taxi ride, a restaurant meal, and an office supply purchase all appear as rows in the same CSV with no expense category attached.
Card sales slips (카드영수증/카드매출전표) are the middle ground. Every credit or debit card transaction generates a paper or digital receipt — sometimes handed to you across the counter, sometimes emailed as a PDF, sometimes only viewable in your banking app as a push notification. Unlike cash receipts, there is no single government portal where all card slips aggregate. Your KB Kookmin card, Shinhan card, and Samsung card each live in separate apps. If you need a unified spreadsheet, you are pulling data from three or four different banking portals.
Simplified receipts (간이영수증) are the wildcard — and for many freelancers, the biggest source of lost deductions. These are handwritten or printed receipts for small transactions under 30,000 won, issued without a business registration number. A corner hardware store, a street-side printer, a mom-and-pop lunch spot — none of them issue 현금영수증. Under Article 160 of the Income Tax Act (소득세법), 간이영수증 are valid expense documentation if they include the date, amount, vendor name, and description of goods or services. But every one of them needs to be manually typed into your spreadsheet — because there is no digital version anywhere.
| Receipt Type | Digital Record? | Main Extraction Challenge | Tax Document Status |
|---|---|---|---|
| Cash receipt (현금영수증) | Yes — NTS HomeTax CSV | No expense category; aggregate only | Full deduction + 소득공제 (Art. 126-2, 조세특례제한법) |
| Card slip (카드영수증) | Partial — per bank app | Scattered across 3-4 banking portals | Full deduction (법인카드) or 소득공제 (개인카드) |
| Simplified receipt (간이영수증) | No — paper only | Handwritten, no digital source at all | Limited: < 30,000 won each, subject to expense ratio cap |
Most freelancers end up with a hybrid approach: download the cash receipt CSV from HomeTax, manually transcribe card slips from banking apps or photos, and type in every 간이영수증 by hand. The result is a spreadsheet built from three completely different workflows — and usually finished the night before the May 31 deadline.
Knowing the three receipt types is step one. The next question is what your tax filing actually demands from each one.
What 종합소득세 Filing Actually Requires From Your Receipts
Comprehensive income tax (종합소득세) is filed once a year, between May 1 and May 31, covering all income and deductible expenses from the previous calendar year. For freelancers and sole proprietors in Korea, this is the single most consequential tax event — and the one where receipt organization determines how much you pay.
Korean tax law gives you two paths for calculating deductible expenses (필요경비). You can use the simplified expense rate (단순경비율), where the NTS applies a flat percentage deduction based on your industry code — no receipts needed. Or you can claim actual expenses (기준경비율), where every won of deduction requires supporting documentation: a receipt, a bank transfer record, or a tax invoice. The actual-expense route typically yields a lower tax bill — but only if your receipt documentation holds up under scrutiny.
This creates the practical workflow that eats freelancers' weekends every spring. Your tax accountant wants one Excel file organized by expense category — Meals (식대), Transport (교통비), Office Supplies (사무용품비), Rent (임차료), Communication (통신비), and so on — with each receipt showing date, vendor, amount, and the receipt type. The NTS audit risk is real: if a tax official flags a deduction and you cannot produce the matching documentation within the statutory 5-year retention period (국세기본법 Article 26-2), the deduction gets disallowed and you owe back taxes plus penalties.
A freelancer with 30 to 50 monthly receipts accumulates 360 to 600 receipts per year. At 3 minutes of manual data entry per receipt — finding the date, typing the vendor name, entering the amount, assigning a category — that is 18 to 30 hours of spreadsheet work compressed into the weeks before May 31. Most of that time isn't spent on tax strategy. It's spent on typing.
The question, then, is how to bypass the typing — and go straight to a categorized spreadsheet.
Step-by-Step: Extracting Receipt Data Into a Categorized Excel
ImageToTable.ai uses Custom Column Extraction: instead of drawing rectangles around fields on a template, you type the column names you want — date, vendor, amount, category — and the AI locates each value by understanding what it means on the page, not where it sits in pixels. The column names you enter become the exact headers of your final spreadsheet. This matters for receipts because no two Korean receipts share the same layout — a Paris Baguette card slip, a handwritten 간이영수증 from a local print shop, and a HomeTax cash receipt printout each arrange the same three data points (date, vendor, amount) in different positions, fonts, and orientations.
Here is the extraction workflow from photo to categorized spreadsheet:
Gather and upload your receipts
Collect all receipt images — photos of paper card slips and 간이영수증 taken with your phone, PDF downloads from banking apps, and screenshots of the HomeTax cash receipt list. The tool accepts JPG, PNG, PDF, and WebP. Upload them all at once: the batch mode merges results into a single spreadsheet with one row per receipt. For a detailed walkthrough of receipt-specific extraction, see how to turn receipt images into an Excel spreadsheet.
Define your extraction columns
Enter the column names that match what your tax accountant needs. A practical column set for Korean receipt extraction:
Transaction Date— 거래일자Vendor Name— 상호명Amount— 금액Receipt Type (options: Cash/현금, Card/카드, Simplified/간이)— receipt type classificationExpense Category (options: Meals/식대, Transport/교통비, Office/사무용품비, Rent/임차료, Communication/통신비, Supplies/소모품비, Other/기타)— expense category inference
The last two columns use Inferred Columns: the AI reads the receipt content and determines both the receipt type and the expense category — even though those words never appear on the receipt itself. A taxi card slip gets classified as Transport, a restaurant cash receipt as Meals, a Daiso 간이영수증 as Supplies.
Review and export your categorized spreadsheet
Click extract. Each receipt produces one row in the output, with date, vendor, amount, receipt type, and expense category all populated. A single-page receipt processes in 5 to 10 seconds. The output downloads as XLSX, CSV, or JSON — already categorized, already sorted. Our accuracy benchmarks confirm up to 99% recognition for printed receipt data under normal photo conditions. If you need item-level detail from multi-item receipts (e.g., a restaurant bill with individual dishes), add columns for individual line items and the tool creates one row per item.
The inferred category column is the feature that transforms this from "faster typing" to "a fundamentally different workflow." Without it, after extraction, you would still need to scan 500 rows and manually label each one as Meals, Transport, or Office — another 2-3 hours of spreadsheet work. With it, the AI assigns the category during extraction based on the vendor name, receipt content, and amount. A 12,000-won transaction at a CU convenience store gets Office/Supplies. A 45,000-won transaction at a BBQ restaurant gets Meals. The categories are reviewable and editable — but you are correcting a few misclassifications, not building the column from scratch.
Try it on your own receipt below — no login required:
Files are processed securely and not stored.
One receipt is straightforward. The real payoff comes when you process a full year at once.
Batch Processing a Year of Receipts Before the May 31 Deadline
The 종합소득세 filing deadline on May 31 is not a soft suggestion — it is the statutory cutoff under Article 70 of the Income Tax Act, and late filing incurs a penalty of 20% of the unpaid tax amount. For freelancers and sole proprietors who have been setting aside tax reserves all year, the weeks before May 31 are not about strategy; they are about documentation assembly.
Batch processing changes the arithmetic. Take the 500-receipt annual workload: at 3 minutes of manual entry per receipt, that is 25 hours — roughly three full working days — compressed into the narrow window between when you close your books for December and when your tax accountant needs the file. At 5 to 10 seconds per receipt with AI extraction, the same 500 receipts process in under an hour. You are not just saving time. You are removing the single largest source of pre-deadline stress from the tax filing calendar.
The batch workflow has a structural advantage over manual entry that is easy to miss on first read. When you manually type 500 receipts into Excel one by one, the categorization step and the data entry step are sequential: you enter all 500 rows, then you go back and categorize them. That second pass nearly doubles the total time. With AI extraction, the inferred category column populates alongside the data — extraction and categorization happen in a single pass. The output is already sorted and ready for your tax accountant.
For freelancers who also deal with tax invoices (세금계산서) alongside receipts — common for those with both B2B and B2C income — the same batch can include both document types with different column definitions. A combined workflow extracting Korean tax invoices and receipts into the same categorized spreadsheet is covered in our guide to Korean tax invoice data extraction. For a broader look at how freelancers prepare receipts for tax season, see how to organize freelancer receipts ahead of tax filing.
A categorized Excel is the output. But the next step is getting it into the tools your tax accountant actually uses.
From Excel to 삼쩜삼, SSEM, or Your Tax Accountant
A categorized Excel spreadsheet of 500 receipts is valuable — but it needs to land somewhere to actually affect your tax filing. The Korean freelancer tax ecosystem has several distinct tools, each with different import expectations.
삼쩜삼 (3.3) is the most widely used tax filing app among Korean freelancers, with over 15 million downloads. It calculates estimated refunds, handles 종합소득세 filing, and connects to the NTS for automatic income data retrieval. What it does not do is receipt data entry — users still need to input expense amounts manually or upload categorized spreadsheets through the app's expense reporting interface. If you walk into your 삼쩜삼 filing session with a pre-categorized Excel of every receipt already sorted by expense type, the data entry step collapses from hours to a copy-paste verification pass.
SSEM (쎔), developed by Viva Republica (the company behind Toss), targets freelancers specifically with receipt management and tax estimation. It includes receipt photo capture with basic OCR, but the recognition is designed for single receipts one at a time — not batch processing. For freelancers with significant receipt volumes, the extraction-to-Excel workflow here fills the gap: batch-process all receipts outside SSEM, then import the categorized file.
자비스 (JOBIS) and CashNote (캐시노트) serve small business owners with broader bookkeeping needs, including sales tracking, expense management, and tax report generation. Both support CSV/Excel import for expenses — making a pre-extracted, categorized spreadsheet the ideal input format.
For freelancers working with a traditional tax accountant (세무사), the workflow is even simpler: you email the categorized Excel file. Most 세무사 offices use Douzone (더존) or similar accounting software internally, and a clean spreadsheet organized by expense category with dates and amounts is exactly the input format their staff expects. No re-entry, no follow-up questions, no "what category is this ₩8,500 transaction from March?"
For freelancers who receive income through both employment (근로소득) and sole proprietorship (사업소득) — common in Korea's platform economy — a single batch can include both the independently contracted work's expense receipts and any additional deductible costs. For extracting payroll-related data alongside receipts, see how to extract Korean payslip data to Excel. For a broader cost comparison of manual versus automated document processing in the Korean market, our market cost comparison for Korean document extraction breaks down the numbers.
Frequently Asked Questions
Can the tool read handwritten 간이영수증?
Yes. The AI's vision model handles handwritten Korean text on receipts — including vendor names written in cursive, hand-written amounts, and stamped receipt forms. Accuracy for handwriting is lower than printed text (expect ~90-95% rather than 99%), and very poor handwriting or faded text may require manual verification. But the alternative — typing every handwritten receipt from scratch — means 100% manual entry, and even 90% extraction is an 18× speed improvement over typing.
How does the receipt type auto-classification work?
The inferred column `Receipt Type (options: Cash/현금, Card/카드, Simplified/간이)` reads visual cues on the receipt itself — the presence of a 현금영수증 approval number, card company branding, or the handwritten 간이영수증 format — and classifies accordingly. It also reads the amount: transactions under 30,000 won without a business registration number are almost certainly 간이영수증. The classification is probabilistic and reviewable; incorrect classifications are quick to override.
Does this replace the need to download my HomeTax 현금영수증 CSV?
It can — but the smarter approach is to use both. The HomeTax CSV gives you an authoritative list of every cash receipt (현금영수증) the NTS already knows about, which is valuable for reconciliation (making sure nothing is missing). But the CSV contains no expense categories and no vendor categorization — it is raw transaction data. The AI extraction workflow adds the categorization layer. For maximum accuracy: download the HomeTax CSV for reconciliation, use AI extraction for categorization and for processing card slips and 간이영수증 that do not appear in the CSV.
What if my receipt uses a mix of Korean and English?
The AI processes mixed-language receipts natively — a Korean vendor name alongside an English product description and Arabic numerals for the amount is a single visual context. Column names can be in Korean, English, or a mix; the AI matches field meaning across languages. This is useful for freelancers with international clients alongside domestic expenses, or businesses that receive both Korean and English-language receipts.
Can I separate personal and business expenses during extraction?
Not automatically — the AI cannot determine your intent for a given receipt. What it can do is add an inferred column like `Business Use (options: Business/업무, Personal/개인, Split/분할)` that flags based on vendor type and transaction amount patterns. The final call on which receipts are business versus personal remains yours, since only you know whether that 7-Eleven receipt was for office supplies or a late-night snack.
How does this compare to Korean receipt-scanning apps like 똑똑 or CashNote?
Apps like 똑똑 (DDOK DDOK) and CashNote (캐시노트) are designed for mobile-first, receipt-by-receipt scanning — take a photo, save it, tag it. They work well for ongoing daily receipt capture. ImageToTable.ai is designed for batch processing: upload 50 or 500 receipts at once and get a single consolidated Excel output. The two approaches are complementary rather than mutually exclusive — capture receipts daily with a scanning app, batch-extract them into a categorized spreadsheet before tax season.
This Year's Tax Filing Does Not Have to End at 2 AM on May 30
The gap between having your receipts and having a spreadsheet your 세무사 can use is not a strategic tax problem. It is a data entry problem — and data entry is exactly the class of task that AI extraction eliminates. Whether you have 50 receipts or 500, the extraction step should take minutes, not the final weekend of May. Upload a receipt — real or sample — and see the date, vendor, amount, and category land in the right columns on the first pass.