What Is Azure AI Document Intelligence?

Azure AI Document Intelligence is Microsoft's cloud service for extracting text, tables, key-value pairs, and structured fields from documents, formerly named Azure Form Recognizer. You reach it through a REST API and SDKs, so it is a developer service you build an application around, not a finished product you sign in and use. Pricing is pay-as-you-go per page on the S0 tier (roughly $1.50 per 1,000 pages for OCR, about $10 for prebuilt and layout, about $30 for custom extraction), plus the engineering time to build the workflow around it. Teams that want a ready-to-use extraction workflow instead of a service to assemble usually compare it against an alternative.

Last updated July 2026.

If you are researching how to pull data out of documents on Microsoft's cloud, Azure AI Document Intelligence is the service you will land on. It shows up in Azure tutorials, in invoice and receipt automation guides, and in nearly every comparison of cloud OCR options. The part that trips people up is what it actually is: a developer service you build on, not a finished product you sign in and use. This article explains what Azure AI Document Intelligence is, what it is used for, how it relates to the old Form Recognizer name, what it costs, where its limits show up, and when teams choose an alternative.

What is Azure AI Document Intelligence?

Azure AI Document Intelligence is a cloud service from Microsoft that extracts text, key-value pairs, tables, and structured fields from documents. It is part of the Azure AI family and was previously called Azure Form Recognizer. You reach it through a REST API and language SDKs, point it at a document, and it returns the recognized content and extracted fields as JSON. It offers ready-made models for common document types, a general model for layout and text, and the option to train your own models on documents you label. It is built for developers who are working inside the Azure platform.

What is Azure AI Document Intelligence used for?

Teams use Azure Document Intelligence to automate data entry from documents: reading invoices, receipts, identity documents, tax forms, and business forms so the values do not have to be keyed by hand. The layout model returns text, tables, and structure from almost any page, while prebuilt and custom models return named fields. Developers call the service from an application and route the JSON it returns into accounting systems, databases, or downstream workflows. Because it is an API rather than an app, the common pattern is to wrap it in software your team builds, which is also where the real work and cost sit.

What is the difference between Azure Document Intelligence and Form Recognizer?

They are the same product under two names. Microsoft renamed Azure Form Recognizer to Azure AI Document Intelligence, so older documentation, SDK package names, and blog posts that mention Form Recognizer describe the service now called Document Intelligence. The capabilities did not shrink in the rename: prebuilt models, the layout and read models, and custom extraction and classification models all carry over. If you find a tutorial that references Form Recognizer, it still applies. The practical takeaway for a buyer is unchanged, that this is a cloud service you integrate into your own application rather than a ready-to-run extraction tool.

What are prebuilt models in Azure Document Intelligence?

Prebuilt models are models Microsoft has already trained for common document types, so you can extract from them without training anything yourself. There are prebuilt models for invoices, receipts, identity documents, US tax forms such as the W-2, health insurance cards, and a general document model, alongside the layout model that returns text and table structure. They are convenient when your documents match a supported type. When they do not, or when you need fields the prebuilt model does not return, you move to a custom model, which means labeling your own documents and training, rather than using the service out of the box.

What is Document Intelligence Studio?

Document Intelligence Studio is Microsoft's web tool for trying the models, labeling documents, and training and testing custom extraction and classification models before you call them from code. It is a useful workbench, but it is a place to build and evaluate models, not a production document workflow. Once a custom model is trained in Studio, you still integrate it into your own application and add the routing, human review, validation, and export steps around it. Studio shortens the model-building part; it does not give you the finished pipeline that turns a stack of mixed documents into clean data in your systems.

Azure Document Intelligence pricing

Azure Document Intelligence pricing is pay-as-you-go per page on the S0 tier: roughly $1.50 per 1,000 pages for the Read (OCR) model, about $10 per 1,000 pages for prebuilt and Layout models, around $3 per 1,000 pages for custom classification, and about $30 per 1,000 pages for custom extraction, with a free F0 tier capped at 500 pages a month. Your real cost depends on which models you run and your monthly volume, since high-volume commitment tiers lower the per-page rate. For the full model-by-model rate table and the costs the headline per-page number hides, see our Azure Document Intelligence pricing breakdown.

How much does Azure Document Intelligence cost?

Azure Document Intelligence pricing is pay-as-you-go by the page, with a free F0 tier that covers up to 500 pages a month but reads only the first two pages of each request, so it suits testing rather than production. On the paid S0 tier the published rates run roughly $1.50 per 1,000 pages for the read (OCR) model, about $10 per 1,000 pages for prebuilt and layout models, around $3 per 1,000 pages for custom classification, and about $30 per 1,000 pages for custom extraction, with neural model training billed by the hour after a free monthly allotment. Always confirm the current numbers on Microsoft's Azure pricing page, since they change.

Those per-page rates are only part of what Azure Document Intelligence costs you. The figure on the pricing page is the recognition step alone. The fuller cost includes the engineering time to build the application around the API, the human review screen, the custom model labeling and training, and the integrations into your systems, which is the spend a ready-to-use product folds into one price. That gap is the thing to weigh when you compare a quoted per-page rate against total cost. See our Azure Document Intelligence alternative for a side-by-side on pricing and workflow.

Is Azure Document Intelligence the same as OCR?

Not exactly. Azure Document Intelligence is built on OCR, but it does more than OCR alone. Its read and layout models recognize printed and handwritten text and detect tables and structure, which is the OCR part. Its prebuilt and custom models add field extraction on top, returning specific values tied to their labels. The distinction matters because OCR returns text while data extraction returns structured fields. A complete workflow also needs classification to sort mixed files and a review step to catch low-confidence reads, and those are pieces you assemble around the service rather than features you switch on.

What are the limitations of Azure Document Intelligence?

Azure Document Intelligence is accurate and scales well, but the limitations teams raise are about assembly rather than recognition quality. It is a service, not an application, so you build the workflow, the human review screen, and the export integrations yourself. Non-standard layouts often need a custom model labeled and trained in Studio before they extract cleanly. You need an Azure subscription and developer and cloud expertise to run it in production. For a single use case or a team without engineering time to spare, that is a meaningful project before the first document is fully processed from input to clean data in your system.

When should you use an Azure Document Intelligence alternative?

An alternative makes sense when you want a finished extraction workflow instead of a service to build on. If you have one use case, a small team, or a timeline in weeks rather than quarters, standing up a cloud project and training custom models can outweigh the benefit. A ready-to-use intelligent document processing product gives you the classify, read, extract, validate, and export workflow the service targets, but as something you sign in and start using. If that fits, see our Azure Document Intelligence alternative for a side-by-side comparison, read how intelligent document processing works end to end, and learn how document classification sorts a mixed batch before extraction runs. Developers can pull the same result through a single OCR API call. For a direct side-by-side with the other major cloud services, see Azure Document Intelligence vs Textract vs Google Document AI, or the dedicated Azure Document Intelligence vs AWS Textract and Azure Document Intelligence vs Google Document AI matchups. If you are comparing options more broadly than the cloud APIs, our roundup of the best OCR software for business weighs the leading tools on accuracy, automation, and price. The surest way to decide is to run one of your own documents through a tool and check the accuracy on your real layouts before committing to a platform.