Google Document AI vs Cloud Vision Pricing for Invoice Extraction

Short answer: Google Cloud Vision and Document AI's Enterprise Document OCR cost exactly the same to read an invoice, $1.50 per 1,000 pages after the first 1,000 free each month. Neither returns invoice fields. The Google product that does, the Document AI Invoice Parser, costs $0.10 per invoice of up to 10 pages, so 1,000 single-page invoices cost $100 instead of $1.50. That 66x jump is the real decision, not Vision against Document AI.

This comparison is for US controllers, AP leads and developers who have already decided to stay inside Google Cloud and need to pick which meter to pay. Every rate below was read from Google's Cloud Vision and Document AI pricing pages in October 2026, in US dollars at standard pay-as-you-go rates. If you would rather see what clean invoice output looks like before writing any code, drop one of your own PDFs into the tool at the top of this page.

Google Document AI vs Cloud Vision pricing side by side

For plain text, the two are priced identically: free for the first 1,000 units a month, $1.50 per 1,000 up to 5 million, then $0.60. The difference is what sits above OCR. Vision stops at text. Document AI adds layout, forms, custom extraction and pretrained invoice parsing, each on its own meter.

Two billing rules change the arithmetic. Vision treats every page of a PDF as a separate image, so a three-page invoice is three units. The Invoice Parser bills in counts, where one count covers up to 10 pages of a single document: 1 to 10 pages is $0.10, 11 to 20 pages is $0.20. And if one PDF holds several invoices, Google says only the first eligible form is processed and charged, so batched scans have to be split before they are sent.

How much does Google Document AI cost per invoice?

The Document AI Invoice Parser costs $0.10 per invoice of up to 10 pages at list price, which is $100 per 1,000 invoices. Text-only OCR on the same invoice is $0.0015 a page. A Custom Extractor costs $0.03 a page, plus a $0.05 an hour hosting fee for each deployed custom processor version.

The Invoice Parser is the closest Google gets to a finished invoice product. It returns supplier, invoice number, dates, totals, tax and line items without training. The catch is the flat $0.10. Most US vendor invoices are one or two pages, so you pay for ten pages and use one. Google offers lower rates on one-year and three-year committed consumption, which helps at scale but not for a team still testing. Our Google Document AI pricing breakdown covers every processor and the commitment tiers.

Is Google Cloud Vision cheaper than Document AI for invoices?

Only if you count the API bill and nothing else. Vision reads an invoice for $1.50 per 1,000 pages, but it returns raw text, so your team has to write and maintain the code that finds the vendor, the invoice number, the total and every line item across hundreds of supplier layouts.

That parsing layer is where Vision projects stall. Regular expressions handle the five biggest vendors and then fall apart on the sixth, a supplier that puts the total in a footer table or splits line descriptions across two rows. Every new vendor becomes an engineering ticket. If the budget comparison does not include developer hours, it is comparing Vision against Document AI on the cheapest 1% of the work. The Cloud Vision API pricing page walks through the Vision meters and that build cost in more detail, and if you are also weighing Amazon, the AWS Textract vs Google Vision comparison runs the same numbers against Textract Analyze Expense at $10 per 1,000 pages.

When the Custom Extractor beats the Invoice Parser

On invoices of one to three pages, the Custom Extractor is cheaper per document than the Invoice Parser: $0.03 a page against a flat $0.10. From four pages up, the Invoice Parser wins, because its $0.10 covers up to ten pages. The break-even sits at about 3.3 pages per invoice.

The per-page saving has two costs attached. A Custom Extractor needs a schema someone defines and maintains, and Google lists a hosting fee of $0.05 an hour for each deployed custom processor version, about $36.50 a month or $438 a year per version left running. At 1,000 single-page invoices a month, the extractor saves $70 on pages and spends about $36.50 on hosting. The saving only becomes meaningful at several thousand short invoices a month.

What 1,000 and 10,000 invoices a month cost on each Google API

At 10,000 single-page invoices a month, Vision or Enterprise OCR costs $13.50, the Custom Extractor about $336.50 with one hosted version, and the Invoice Parser $1,000. Only the last two return invoice fields without code of your own.

For comparison, a flat extractor that returns header fields and every line item to Excel, CSV or an accounting import costs $49 a month for 2,500 pages on InvoiceExtractor Starter and $149 a month for 10,000 pages on Plus, with no Google Cloud project, no schema to maintain and no hosting meter. At 10,000 invoices a month that is $149 against $1,000 on the Invoice Parser. Line items are included on both plans, which is the part that usually forces teams off text-only OCR in the first place. See invoice line item extraction for what that output looks like.

Is Google Document AI free?

Partly. Enterprise Document OCR includes the first 1,000 pages a month at no charge, the same allowance Cloud Vision gives its text detection features. The Invoice Parser, Form Parser, Layout Parser and Custom Extractor have no free monthly allowance on the pricing page, though new Google Cloud accounts receive trial credits.

Google also states that failed requests returning 4xx or 5xx codes are not billed, and that a document with no eligible form is not processed or charged. That makes small tests cheap. It does not make production cheap, because a successful parse of a one-page invoice is still $0.10.

Which Google API should an AP team pay for?

Pay for Cloud Vision or Enterprise OCR only if you need searchable text, not data. Pay for the Invoice Parser if you need fields and line items with no training and your volume is modest. Consider the Custom Extractor only above a few thousand short invoices a month with an engineer to own the schema.

One adjacent case comes up often in the same evaluation. Teams pricing the Layout Parser are frequently trying to answer questions from long contracts, vendor agreements or policies rather than pull invoice totals. That is a retrieval job, and a tool built to chat with PDF contracts and reports is a better fit than wiring Layout Parser chunks into your own pipeline. For invoices themselves, the choice comes down to how much of the work you want to own. If the answer is none of it, the Google Document AI alternatives page compares the tools that return finished rows, and the converter at the top of this page will show you the output on your own invoice in under a minute.