Does Amazon Textract Read Receipts and Invoices?

Yes, Amazon Textract reads receipts and invoices. It does it through a dedicated API called AnalyzeExpense, which is separate from Textract's plain-text and general document APIs. AnalyzeExpense returns the fields that matter on a bill: vendor, date, invoice or receipt number, subtotal, tax, total, and the line items, each with a confidence score. The catch is that Textract is a raw AWS building block, not a finished tool: you get JSON back, you write the code, and there is no Excel or CSV button.

Last updated July 2026. Every price, field name, and format below comes from Amazon's own Textract documentation and pricing page, cited inline.

Does Amazon Textract work for receipts and invoices?

Yes, through the AnalyzeExpense API, which AWS built specifically for invoices and receipts. It extracts the date, the receipt or invoice number, item prices, the total, and payment terms, and it can identify the vendor even when the name appears only in a logo. AnalyzeExpense also normalizes the many names a field can go by, so a "balance due," an "amount due," and a "total" all map to a single standard label. That normalization is the part that makes it more than raw OCR.

What is the difference between AnalyzeExpense and AnalyzeDocument?

AnalyzeExpense is purpose-built for receipts and invoices and returns labeled expense fields. AnalyzeDocument is the general-purpose API, and you turn on features individually: Tables pulls grid structure, Forms pulls key-value pairs, Queries lets you ask for a field in plain language, and Signatures detects signatures. For a receipt, AnalyzeExpense is the right call because it already knows what a vendor and a total are. Using AnalyzeDocument Forms on a receipt gives you raw key-value pairs that you then have to interpret yourself, and it costs far more per page.

How much does Amazon Textract cost per page?

AnalyzeExpense costs $0.01 per page for the first million pages a month, dropping to $0.008 per page after that, in the US West (Oregon) region. Other APIs are priced separately. The table below lists the first-tier price per 1,000 pages so the APIs are easy to compare. Prices vary by AWS region, so confirm your own region on the pricing page before you budget.

Is Amazon Textract free? Does it have a free tier?

There is a free tier, but it is a trial, not a permanent free plan. New AWS customers get 100 pages a month of AnalyzeExpense free for the first three months. After that, or above 100 pages, you pay the per-page rate. There is no free forever option, and there is no way to use Textract without an AWS account, so the real entry cost is the AWS setup, not the first hundred pages.

What data fields can Textract extract from an invoice?

AnalyzeExpense returns two groups of fields. Summary fields describe the whole document, and line-item fields describe each row. Every returned value carries a Confidence score from 0 to 100 and a Geometry bounding box telling you where on the page it was found, which is what you use to flag low-confidence values for review.

Does Amazon Textract extract line items?

Yes. AnalyzeExpense returns line-item groups with item description, quantity, price, unit price, and product code, alongside the summary total. This is one of its genuine strengths and it is more than most built-in accounting scanners give you, since tools like QuickBooks Online and Wave stop at the header total. The work you still own is stitching those line-item groups into the rows your spreadsheet or ledger expects.

What file formats does Amazon Textract support?

The synchronous API takes image bytes, meaning JPEG and PNG. The asynchronous API, StartExpenseAnalysis, accepts JPEG, PNG, TIFF, and PDF, which is what you use for multipage documents and larger batches. So a multipage PDF invoice goes through the async path, and a single phone photo of a receipt can go through either.

What languages does Textract support for receipts?

This is the limit most teams miss. Textract handles printed text, forms, and tables in English, German, French, Spanish, Italian, and Portuguese, but AWS states plainly that handwriting, invoices and receipts, identity documents, and Queries processing are English only. If your receipts are in Spanish or French, AnalyzeExpense is not built for them, and that constraint is easy to discover only after you have written the integration.

Do I need to write code to use Amazon Textract?

Yes. Textract is an API, not an application. There is no upload page that returns a spreadsheet. You create an AWS account, configure IAM permissions, call the API with the SDK, parse the JSON response, handle the async job polling for PDFs, and build your own export to Excel or CSV. For an engineering team that wants receipt data flowing into its own software, that control is the point. For a bookkeeper who wants a spreadsheet this afternoon, it is a project, not a tool. If the documents arrive as email attachments in the first place, you also need a way to pull them out of the inbox before any of this runs.

Textract or a done-for-you extractor?

Choose based on what you are building. If you are wiring receipt data into an application and you have engineers, a raw receipt OCR API or invoice OCR API is the honest comparison to make, and the questions to weigh are export format, language coverage, and how much JSON parsing you want to own. If you want to drop a batch of receipts in and get an Excel file back with no AWS account and no code, that is a done-for-you tool, and our AI invoice data extraction is built for exactly that. Textract is excellent at the extraction; whether you want to build the rest around it is the real decision.