Azure AI Document Intelligence, the service formerly called Form Recognizer, has prebuilt receipt and invoice models that cost $0.01 a page and work well. What it does not have is an export. It returns JSON, it needs an Azure subscription, a provisioned resource, an endpoint, a key, and code to call it, and the Document Intelligence Studio is a developer test harness rather than a way to turn a folder of receipts into a spreadsheet. ReceiptOCR reads the same documents in a browser, shows every field for review, and exports Excel, CSV, or JSON. Upload a receipt below.
Upload your receipts and invoices
Drop files here or click to upload
Up to 50 files
Uploading...
Azure Document Intelligence is a serious extraction service and the per-page price is among the lowest in the market. The cost that does not appear on the pricing page is everything you must build before a single receipt reaches a spreadsheet.
The prebuilt receipt and invoice models return structured JSON. There is no spreadsheet export anywhere in the product. XLSX is an input format for Layout, never an output, so turning results into rows is code you write.
Before the first document, you create an Azure account, provision a Document Intelligence resource, and collect an endpoint and a key from the portal. That is a cloud onboarding project, not a signup.
Document Intelligence Studio lets you drop in a file and eyeball the fields, which is useful for testing. It is a developer harness, not a way for a bookkeeper to process a month of receipts and download a sheet.
The F0 free tier gives 500 pages a month but analyzes a maximum of two pages per document, with a 4 MB file limit. Any real multi-page invoice pushes you onto the paid S0 tier immediately.
ReceiptOCR returns what the prebuilt receipt and invoice models return, except a person can operate it and the output is a spreadsheet. No subscription, no resource to provision, no parsing code.
Download extracted receipts and invoices as a spreadsheet, a CSV, or a QuickBooks-ready file. No code stands between the document and the row in your books.
No Azure account, no resource, no endpoint, no key rotation. Open a browser, upload the documents, take the file.
Every extracted field is on screen and editable before the data leaves. A confidence score in a payload tells you something might be wrong; it does not fix it.
Multi-page invoices go through whole. There is no free-tier page cap to design around and no 4 MB file limit to fight.
Merchant, date, subtotal, tax, tip, and total, plus each line with description, quantity, and unit price where the document carries them.
The same fields come back as JSON over a REST endpoint. You get the developer path without being forced onto it from day one.
From a document to a spreadsheet without provisioning anything.
Receipts, invoices, or both in the same batch. PDFs, scans, and phone photos all go in together, and you do not pick a model first.
Tip: Include a multi-page invoice. Per-page billing and a two-page free-tier cap are exactly where the cost math between these tools diverges.
Review merchant, date, tax, total, and every line item on screen. Correct anything the engine flagged before it becomes a row in your books.
Download Excel, CSV, or a QuickBooks-ready file, or call the REST API and receive the same fields as JSON inside your own workflow.
Built for US teams that need receipt and invoice data in a spreadsheet, whether or not anyone on staff can provision an Azure resource.
Client documents have to become a clean sheet this week. An endpoint, a key, and a JSON sample do not get you there; a browser upload and an Excel download do.
Supplier invoices in dozens of layouts headed for a ledger. The value is the fields, not the resource, and nobody in AP is opening the Azure portal.
You could wire up Document Intelligence. First you want to see the extraction quality on your own documents without spending a day on setup to find out.
A few hundred documents a month, no cloud engineer, and no appetite for a subscription whose bill you have to monitor.
Last updated July 2026.
Azure AI Document Intelligence is Microsoft's document extraction service, and it is the product that used to be called Form Recognizer. Its prebuilt receipt and invoice models return structured fields, including line items, for $0.01 a page. The extraction is good and the price is low. The question this page answers is whether you need a cloud service, or whether you need the data a cloud service would give you, in a spreadsheet, without provisioning anything.
Yes, it is the same service. Microsoft renamed Form Recognizer to Azure AI Document Intelligence, so the two names describe one product and older documentation and tutorials still use the old name. It offers prebuilt models, including prebuilt-receipt and prebuilt-invoice, a Read model for plain OCR, a Layout model for tables and structure, and custom models you train yourself. You reach it through a REST API or the C#, Python, Java, and JavaScript SDKs, or you test it by hand in the browser-based Document Intelligence Studio.
These figures come from Microsoft's own pricing for the pay-as-you-go S0 tier in July 2026. Confirm on the Azure pricing page before you commit, because cloud prices move and vary by region.
| Model | What it returns | Price per 1,000 pages |
|---|---|---|
| Prebuilt (receipt, invoice, ID, W-2) | Named fields plus line items, as JSON | $10.00 ($0.01 per page) |
| Read | Raw text only, no fields | $1.50 |
| Custom extraction | Fields from a model you train | $30.00 |
| Free tier (F0) | Max 2 pages per document, 4 MB files | 500 pages per month free |
Commitment tiers exist for volume, and Microsoft has published prebuilt blocks such as 20,000 pages for $190 a month and 100,000 for $900. Two things decide your real cost. Billing is per page, not per document, so a four-page invoice costs four times what a receipt costs. And the sticker price excludes the part you actually have to pay for: the engineering time to call the API, parse the JSON, handle failures, and build somewhere for a human to correct a misread total.
No. It returns JSON, and there is no spreadsheet export anywhere in the product. This trips people up because XLSX appears in the documentation as a supported input format for the Read and Layout models, which is the opposite of an export. If your goal is a spreadsheet, Azure gives you an extraction engine and leaves the last mile to you: someone writes the code that turns the JSON entities into rows, and someone maintains it.
The prebuilt-receipt model returns a structured JSON payload with merchant name, phone, and address, transaction date and time, subtotal, tax, tip, total, and a receipt type, along with an Items array carrying description, quantity, unit price, and total price for each line. The prebuilt-invoice model does the equivalent for supplier bills. The field coverage is genuinely strong, which is exactly why the missing export is so frustrating: the data you want is right there in a format nobody in accounting can open.
Yes. You need an Azure subscription, a Document Intelligence resource provisioned in the portal, and the endpoint and key that resource issues. Then you call the service from code, or you paste single files into Document Intelligence Studio to inspect the results. There is no consumer signup and no application for a non-developer. If your organization already runs on Azure, this is a small step. If it does not, you are adopting a cloud platform in order to read a receipt.
For testing, yes. For work, rarely. The F0 tier includes 500 pages a month, but it analyzes only the first two pages of any document and caps files at 4 MB, with a one-transaction-per-second limit. A two-page cap is fine for receipts and useless for the multi-page invoices that make up most real accounts payable, so any serious volume moves to the paid S0 tier on the first real document.
These tools overlap on the extraction and diverge on who is meant to operate them. Neither wins everywhere.
| Capability | Azure Document Intelligence | ReceiptOCR |
|---|---|---|
| How you use it | Provision a resource, call from code | Upload in a browser, or call the API |
| Excel or CSV export | No, JSON only | Yes, Excel, CSV, and QuickBooks |
| Account required | Azure subscription and resource | A browser |
| Pricing unit | Per page, $0.01 prebuilt | By document volume |
| Custom model training | Yes, a genuine strength | No, prebuilt for receipts and invoices |
| Review before export | You build it | Built in |
| Who operates it | Developers and cloud engineers | Anyone, plus developers via the API |
Choose it when you are building on Azure. If your stack already lives there, your identity and compliance posture is already Microsoft, and you have engineers who will embed extraction into an application, Document Intelligence is an excellent service. The prebuilt models are strong, custom model training is a real capability that prebuilt-only tools cannot match, and at $0.01 a page the marginal cost at scale is hard to argue with. We are not going to pretend otherwise.
Choose it, too, when extraction is one component of a system you already own. If you have the ingestion, the queueing, and the human review interface, and you only need a good model behind an endpoint, a cloud API fits that architecture better than a finished application would.
If nobody is going to provision a resource, the cloud service is the part you can skip. For expense receipts, receipt OCR software reads them in a browser and the receipt to Excel converter lands them in a spreadsheet. For vendor bills, invoice OCR software pulls header fields and line items, and the invoice PDF to Excel converter answers the PDF conversion job directly. Large batches run through the bulk receipt scanner. When you do want to build, the receipt OCR API and invoice OCR API return the same fields as JSON. Comparing the other big clouds? The AWS Textract alternative and Google Document AI alternative pages cover per-page billing on each, Google Vision OCR covers the raw-text tier below them, and the intelligent document processing overview maps the whole category. The OCR software pillar explains where each type fits.
Yes. Microsoft renamed Azure Form Recognizer to Azure AI Document Intelligence, so the two names refer to one service. Older tutorials and blog posts still say Form Recognizer, and the prebuilt receipt and invoice models are the same models under either name.
On the pay-as-you-go S0 tier in July 2026, the prebuilt models, including receipt and invoice, cost $10 per 1,000 pages, which is $0.01 a page. Read, which returns raw text only, is $1.50 per 1,000 pages, and custom extraction is $30. Billing is per page, so a four-page invoice costs four times a one-page receipt.
No. It returns JSON only. XLSX appears in the docs as a supported input format for Read and Layout, never as an output. Converting the JSON into spreadsheet rows is code you write and maintain, which is the main gap when the deliverable is a sheet.
Yes. You need an Azure subscription, a provisioned Document Intelligence resource, and its endpoint and key, then you call the service from code. Document Intelligence Studio lets you test single files in a browser, but it is a developer harness, not a bulk receipts-to-spreadsheet tool.
For testing, yes. For work, rarely. The F0 free tier gives 500 pages a month but analyzes a maximum of two pages per document and caps files at 4 MB. Any real multi-page invoice exceeds that, so production volume moves to the paid S0 tier almost immediately.
A tool that does the same receipt and invoice extraction but runs in a browser and exports a spreadsheet. ReceiptOCR reads unfamiliar layouts, shows every field for review, exports Excel, CSV, JSON, or a QuickBooks-ready file, and offers a REST API when you want one, with no Azure subscription to provision.
They are closely matched on the headline rate. Azure prebuilt models and AWS Textract AnalyzeExpense both run about $0.01 a page, and Google Document AI Invoice and Expense parsers are $10 per 1,000 pages, also $0.01. None of the three exports Excel or CSV, and all three require a cloud account and code, so the deciding factor is usually which cloud you already run on.
Amazon's cloud OCR API, compared honestly.
Google's Invoice and Expense parsers, compared.
The raw-text tier below Document AI, explained.
The same extracted fields as structured JSON.
Turn a pile of receipts into a clean spreadsheet.
The three types of OCR, and which one you need.
