Google Document AI Page Limit Exceeded: Why It Happens and How to Fix It

Last updated July 2026.

The Google Document AI "page limit exceeded" error means your document is over the online processing limit of 15 pages. Online (synchronous) processing accepts up to 15 pages per document, or 30 with imageless mode, and up to 40 MB. To process a longer document you switch to batch (asynchronous) processing, which handles up to 1,000 pages per file, files up to 1 GB, and up to 5,000 files per batch request. The fix is not a quota increase, it is a different endpoint.

This is the first wall most teams hit with Document AI. You wire up online processing, it works beautifully on a two-page test file, and then a real 22-page contract returns the page limit error. The numbers below are read from Google's Document AI limits and quotas documentation in July 2026. Google updates these pages, so confirm the current figures before you build around them.

Why does Google Document AI say page limit exceeded?

Because you sent more pages than online processing allows. The synchronous process method, the one that returns a result in the same call, caps at 15 pages per document. When your file has more pages than that, the API rejects it with a page limit error rather than truncating it. The same thing happens if the file is over 40 MB. It is a hard content limit, not a soft one you can pay to raise, so the answer is always to reduce the page count in the request or move to the asynchronous path.

What is the Google Document AI page limit?

Online processing is capped at 15 pages and 40 MB per document. Turning on imageless mode raises the page cap to 30, because it strips the rendered page images from the response and processes text only. Batch processing raises the ceiling much higher: up to 1,000 pages per file for most processors, a file-size limit of 1 GB, and up to 5,000 files in a single batch request. So the effective page limit is 15 online, 30 online imageless, and up to 1,000 in batch.

Document AI limits at a glance

See how these ceilings compare with Azure and AWS Textract on the OCR API limits comparison.

How do I process more than 15 pages in Document AI?

Use batch processing. Instead of the synchronous process method, you call the asynchronous batchProcess method, point it at your files in a Cloud Storage bucket, and Document AI writes the results back to another bucket when the operation finishes. Batch handles up to 1,000 pages per file and 5,000 files per request, so a long contract or a stack of statements goes through in one operation. The cost is the same shape as every async OCR path: a Cloud Storage bucket, a long-running operation to wait on, and result files to read and reassemble.

If splitting is easier for your workflow, you can also break a document into chunks of 15 pages or fewer and send each through online processing, then merge the results by page number. That keeps the synchronous simplicity but adds the splitting and merging logic, and it multiplies your request count, which matters against your per-minute quota.

What is imageless mode in Document AI?

Imageless mode is an online processing option that skips returning the rendered page images and processes the text layer only. Because the response is lighter, Google raises the online page cap from 15 to 30 when you use it. It is a useful middle ground when your documents are 16 to 30 pages and you want to stay on the synchronous path, but it does not help beyond 30 pages, and you lose the page image in the response. For anything over 30 pages, batch is the only option.

What is the batch processing limit for Document AI?

Batch processing accepts up to 1,000 pages per file for most processors, files up to 1 GB, and up to 5,000 files per batch request, with a limit of 5 concurrent batch requests per project. There is also a throughput quota measured in pages per minute that varies by processor and whether you have provisioned capacity. If you run large nightly jobs, watch the concurrent-request and pages-per-minute quotas as closely as the per-file page limit, because those are what throttle a big backfill.

Does the page limit differ by processor?

Yes. The per-file batch page limit varies by processor. Some, like a custom splitter, allow the full 1,000 pages, while form-style processors and prebuilt parsers can cap lower. The online 15-page limit is consistent across processors, but always check the specific processor you are using, because a prebuilt parser such as the W2 or invoice parser may enforce a tighter page limit than Enterprise Document OCR. When in doubt, the processor's own documentation lists its exact content limits.

Is there a way to avoid the page limit entirely?

If you would rather not build batch processing, Cloud Storage handling and result reassembly, a ready-to-use product removes the page limit from your code. You upload the full document in one step and it batches, splits and reassembles behind the scenes, then returns fields with a review step for low-confidence values. That is how DocuOCR works, and it means a 300-page file is one upload rather than a batch job you maintain. If the documents you are processing are invoices, you can also just pull the line items out of an invoice directly, without standing up a Document AI pipeline for it.

The bottom line on the Document AI page limit

The Google Document AI page limit exceeded error is not a bug or a quota you need to raise; it is the 15-page online cap doing its job. Move to batch processing for up to 1,000 pages, or turn on imageless mode for up to 30, or split your document into 15-page chunks. Whichever you choose, design for your longest document, because the limit shapes your architecture even when your average file is short. To see the full cross-vendor picture, read the OCR API limits comparison, and for what Document AI costs per 1,000 pages, the OCR API pricing comparison.