Google Document AI Limits: Page Limit, File Size and Quotas

See how these ceilings compare with Azure and AWS on the OCR API limits comparison, and what Document AI costs per 1,000 pages on the Google Document AI pricing breakdown.

What is the Google Document AI page limit?

For online (synchronous) processing the page limit is 15 pages per request, or 30 pages if you enable imageless mode. For batch (asynchronous) processing the limit is set per file and varies by processor, from 100 up to 1,000 pages. So the practical answer is 15 pages online and up to 1,000 pages in batch. The 15-page online cap is exactly what triggers the PAGE_LIMIT_EXCEEDED error when you send a longer document to the synchronous endpoint.

Why does Google Document AI say page limit exceeded?

You sent more than 15 pages to the online (synchronous) processor. Online ProcessDocument accepts 15 pages by default. To fix it you have three options: enable imageless mode to raise the online cap to 30 pages, split the file into 15-page chunks, or switch to batch processing, which reads up to 1,000 pages per file from Cloud Storage. Most teams move to batch once documents routinely run past 15 pages.

What is the Google Document AI file size limit?

Online processing accepts a file up to 40 MB. Batch processing accepts up to 1 GB per file, read from a Cloud Storage bucket, with up to 5,000 files in a single batch request. So a large scanned PDF has to go through batch, not the online endpoint. Images are also capped at 40 megapixels per page. Files over these sizes have to be split or downsampled before Document AI will accept them.

How many pages can Google Document AI process in batch?

In batch mode the maximum pages per file depends on the processor, from 100 pages up to 1,000. Enterprise Document OCR reads up to 1,000 pages per file, while some prebuilt parsers cap lower, so check the limit for your specific processor. A single batch request can include up to 5,000 files, and you can run 5 concurrent batch requests per project. Batch reads files from a Cloud Storage bucket and writes results back to one.

What is imageless mode in Google Document AI?

Imageless mode is an option on ProcessRequest that raises the online page limit from 15 to 30 pages. It works only when you process pages contiguously starting from page 1, and it skips returning the rendered page images, which is what lets it handle more pages in a synchronous call. It is a quick way to double the online ceiling without moving to batch, but 30 pages is still the cap, so longer documents still need batch.

Does Google Document AI have a rate limit?

Yes. Document AI enforces quotas you can view and adjust in the Google Cloud console, including 5 concurrent batch requests per project and provisioned throughput of about 120 pages per minute for base processor versions and 30 pages per minute for Pro base versions. These quotas are adjustable, unlike the content limits such as the 15-page online cap and the 1 GB batch file size, which are system limits that cannot be changed.

Online vs batch processing in Document AI, which do I use?

Use online (synchronous) processing for short documents of 15 pages or fewer when you want an immediate response. Use batch (asynchronous) processing for anything longer, for high volume, or for files up to 1 GB, since it reads up to 1,000 pages per file from Cloud Storage and handles 5,000 files per request. The tradeoff is that batch is asynchronous: you submit the job, then poll or wait for the operation to finish, rather than getting the result inline.

How do I process a document longer than 15 pages?

You have three paths. Turn on imageless mode to reach 30 pages online. Split the document into 15-page chunks and process each online, then merge the results by page number. Or send the whole file to batch processing, which takes up to 1,000 pages per file. Batch is the usual choice for documents that routinely exceed 15 pages, because splitting and stitching adds code and can break tables that span a page boundary.

What file formats does Google Document AI support?

Document AI reads PDF, TIFF and GIF, plus common image types such as JPEG, PNG, BMP and WEBP for online processing, with results depending on the processor. PDF and TIFF are the formats used for multi-page documents. The 40-megapixel-per-page image resolution limit applies across formats. For a Word or Excel file you convert to PDF first, since Document AI processes document images and PDFs rather than native Office files.

Is Google Document AI cheaper than the page limits suggest?

The limits and the price are separate questions. Enterprise Document OCR is about $1.50 per 1,000 pages, but the prebuilt Invoice, Expense and Utility parsers bill in 10-page blocks at roughly $0.10 per block, so a 1-page invoice consumes a full block. The 15-page online cap does not change the rate, but the block billing on prebuilt parsers can make short documents cost more than the page count implies. See the pricing breakdown for the per-processor rates.