The short answer: Google Document AI charges per page, per processor, on your Google Cloud bill, with no plan and no seat fee. Enterprise Document OCR is $1.50 per 1,000 pages after the first 1,000 pages a month. Layout Parser is $10 per 1,000 pages. Form Parser and Custom Extractor are $30 per 1,000 pages up to one million pages a month, then $20. Specialized parsers bill per document instead: $0.10 per ten pages for invoices and receipts, $0.75 for each bank statement.
Those are the list prices on Google's own Document AI pricing page, in USD, pay as you go. They are low, and that is the honest headline. What catches budget owners out is which processor they actually need for the job, how a "page" is counted, the hosting fee on custom models, and the cost of everything that sits between the API and a finished spreadsheet. This guide covers all of it.
Google Document AI pricing table
Each processor has its own meter. You create a processor in a Google Cloud project, send pages to it, and pay that processor's rate. There is nothing to buy up front.
| Processor | Price (USD) | Per page | What it returns |
|---|---|---|---|
| Enterprise Document OCR | First 1,000 pages a month at no charge, $1.50 per 1,000 to 5 million, $0.60 per 1,000 above | $0.0015 | Text, layout blocks, reading order. No table structure. |
| OCR add ons | $6.00 per 1,000 pages | $0.006 | Extra OCR features such as math and font style detection |
| Layout Parser | $10.00 per 1,000 pages | $0.01 | Document structure, tables, headings and chunks for search or RAG |
| Form Parser | $30.00 per 1,000 pages to 1 million, $20.00 above | $0.03 | Tables, key and value pairs, checkboxes |
| Custom Extractor | $30.00 per 1,000 pages to 1 million, $20.00 above | $0.03 | The fields you define, generative or trained |
| Custom Classifier, Custom Splitter | $5.00 per 1,000 pages to 1 million, $3.00 above | $0.005 | Document type labels and split points |
Google does not bill failed requests (4xx or 5xx responses). If you pay in a currency other than USD, the prices on your Cloud Platform SKUs apply instead of these.
What do the specialized parsers cost?
The pretrained parsers are billed differently. Most charge per "count", and one count covers up to ten pages of a single document. So a one page invoice and a ten page invoice both cost $0.10, and an eleven page invoice costs $0.20.
| Pretrained parser | Price |
|---|---|
| Invoice parser | $0.10 per count (up to 10 pages) |
| Expense parser (receipts) | $0.10 per count |
| Utility parser | $0.10 per count, limited access |
| US driver license, US passport, identity proofing | $0.10 per count |
| Bank statement parser | $0.75 per classified document |
| Pay slip parser, W2 parser | $0.30 per classified document |
| Lending and procurement splitter and classifier | $0.05 per classified document |
If a document holds several eligible forms, only the first is processed and charged. If it holds none, nothing is processed and nothing is charged. Some of these parsers, including the utility parser and the procurement classifier, are only available to customers Google approves through a limited access form, so check availability before you plan around a price.
How does Google Document AI count a page?
For PDFs, every page in the file is one page, and each image in a TIFF is one page. Other formats are counted by size: a Word file or an HTML file is one page per 3,000 characters, each Excel tab is one page, and each PowerPoint slide is one page. That matters for Layout Parser, which accepts office formats, because a long Word contract can count as many more pages than it prints on.
Processing limits shape the bill too. Form Parser, Layout Parser and OCR accept at most 15 pages per online (synchronous) request, or 30 in imageless mode. Longer files go through batch processing, which allows 100 pages per document on Form Parser and 500 on Layout Parser and OCR, with files up to 40 MB online and 1 GB in batch. Batch jobs read from and write to Cloud Storage, which is a small separate line on the same bill.
What Google Document AI costs on real monthly workloads
Here is the list price bill for five workloads a finance team might send through the two processors that return tables. These are API charges only.
| Monthly workload | Pages | Layout Parser | Form Parser |
|---|---|---|---|
| 40 one page vendor invoices | 40 | $0.40 | $1.20 |
| 45 twelve page bank statements | 540 | $5.40 | $16.20 |
| 150 thirty page brokerage statements | 4,500 | $45.00 | $135.00 |
| 300 twenty page loan files | 6,000 | $60.00 | $180.00 |
| 2,000 one page receipts | 2,000 | $20.00 | $60.00 |
Two things stand out. First, the processor you pick triples or cuts the bill: Form Parser costs three times Layout Parser for the same pages, so test both on your documents before assuming you need the more expensive one. Second, the bank statement parser at $0.75 a document is cheaper than Form Parser on long statements (a 30 page statement is $0.90 on Form Parser) and dearer on short ones, but it is limited access, so most teams cannot count on it.
What costs are not on the pricing page?
The rate card is the smallest part of most Document AI budgets. These are the lines that show up later.
- Custom processor hosting. A deployed custom extractor or classifier version costs $0.05 an hour to host whether or not you send it anything. Google's own example puts one version deployed for a year at $438.
- Engineering. Document AI returns a Document object in JSON. Table cells come with text anchors into the full text, not as rows ready for Excel. Someone writes the code that rebuilds rows, keeps amounts numeric, handles merged headers and negatives in parentheses, and splits files over the page limit. The Document AI Toolbox library for Python can turn a parsed table into a pandas DataFrame, which helps, but the pipeline around it is still yours to build and maintain.
- Setup and access. A Google Cloud project with billing, the API enabled, a processor per document type, a service account and, for batch, a storage bucket. In many companies each of those needs an IT or security ticket.
- Monitoring spend. Per page meters are easy to run up by accident, for example by pointing a batch job at the wrong folder. Teams that run several cloud APIs usually put budget alerts or cloud cost monitoring that watches Google Cloud and SaaS spend in one place on the project before it goes live.
Is Google Document AI free?
Only a slice is. The first 1,000 pages a month of Enterprise Document OCR are listed at no charge, but OCR returns text, not tables. Layout Parser and Form Parser, the processors that extract tables, bill from the first page. New Google Cloud accounts receive trial credits that can be spent on Document AI, which is useful for a proof of concept but not a way to run a monthly process.
Are there discounts for committed use?
Yes. Google lists two savings plans next to the pay as you go column. A one year commitment takes 10 percent off (Form Parser drops to $27 per 1,000 pages, Layout Parser to $9) and a three year commitment takes 20 percent off ($24 and $8). You may need to opt in for the discount to apply. The per document prices on the bank statement, pay slip and W2 parsers have no savings plan price listed. For very large volumes Google also offers custom quotes through sales.
Is Google Document AI cheaper than Amazon Textract and Azure?
For table extraction the three clouds land close together, and the difference comes down to which processor or feature you call.
| Capability | Google Document AI | Amazon Textract | Azure Document Intelligence |
|---|---|---|---|
| Plain OCR, per 1,000 pages | $1.50 | $1.50 | $1.50 |
| Tables and layout, per 1,000 pages | $10 (Layout Parser) | $15 (Tables), $4 (Layout without table cells) | $10 (Layout) |
| Tables plus key and value pairs | $30 (Form Parser) | $65 (Tables and Forms together) | $30 (custom extraction model) |
| Custom extraction | $30 plus hosting | Queries at $15, adapters extra | $30 |
The full rate cards are in our Amazon Textract pricing and Azure Document Intelligence pricing guides. In practice, teams pick the cloud they already run on, because the setup, identity and billing work is already done there.
When is a per file converter cheaper than Document AI?
On the API bill alone, rarely. Document AI reads pages for one to three cents each, and no subscription product will beat that on raw volume. The comparison changes when you count the people. If the numbers need to end up in Excel and the people who need them are accountants, underwriters or analysts rather than developers, the real price of Document AI is the pipeline, and a single developer day costs more than a year of a converter plan.
That is the case pdfxlsx is built for. You upload a PDF in the browser, scanned pages go through OCR automatically, and every table comes back on its own sheet as XLSX or CSV. Plans are billed per file, not per page: $49 a month for 50 files, $99 for 500, and $199 for unlimited files with a REST API that returns the workbook itself. A 40 page statement counts once. Our Google Document AI alternative page puts the two side by side on five workloads, including the rows where Google is cheaper, and the PDF to Excel API page shows the three endpoints for teams that still want to call it from code.
How much does Google Document AI cost for 1,000 pages?
At list price, 1,000 pages cost $1.50 on Enterprise Document OCR (or nothing, if they are your first 1,000 of the month), $10 on Layout Parser and $30 on Form Parser or Custom Extractor. For invoices on the invoice parser, the price depends on documents rather than pages: 1,000 one page invoices cost $100, while 100 ten page invoices cost $10.
Is Google Document AI worth it?
If you are a developer building document reading into a product or an internal system on Google Cloud, yes. The per page rates are among the lowest available, the processor family covers classification, splitting and custom fields, and your data stays inside your own cloud account. If what you need is last month's statements and reports in a spreadsheet by Friday, try the converter at the top of this page on your hardest PDF first, then compare the sheet with what a Document AI prototype would take to build. The plans are on our pricing page.