---
title: AI Document Analysis
description: A new endpoint reads a KYC document with AI and returns an approval-rate signal, so you can pre-screen files before submitting them for verification.
date: 2026-06-14
category: New Feature
categoryType: feature
isChangelog: true
---

You can now pre-screen KYC documents before submitting them. The new `POST /v1/upload/analyze` endpoint reads a PDF, JPG, or PNG with AI, checks it against the rules for its type, and returns an approval-rate signal with a short reason.

<!--more-->

## TL;DR

- New endpoint: `POST /v1/upload/analyze`
- Returns `approval_rate` of `low`, `medium`, or `high` plus a one-line `description`
- Supports 13 document types, from identity and passport to bank and financial statements
- Optional `metadata` cross-checks values you already collected against the document

## How it works

Send a `file` (PDF, JPG, or PNG up to 5MB) and a `type`. The analysis runs the document against the rule set for that type and returns how confident it is that the document should be approved, with a short reason you can show to the customer or your reviewers.

The result is a signal, not a decision. KYC approval still comes from the verification flow; this lets you catch an expired ID or an out-of-date proof of address early, before submitting.

## Optional metadata

Pass `metadata` as a JSON string of values you already have, such as `full_name`, `date_of_birth`, or `address`, and they are matched against the document. A clear mismatch caps the rating at `low` and the reason explains it. Requested transaction limits (`requested_per_transaction`, `requested_daily`, `requested_monthly`) are checked for sufficient capacity instead of an exact match.

## Get started

See the [Analyze Document documentation](/docs/essentials/analyze-document). The full request and response schema is in the [BlindPay API Docs](https://api.blindpay.com/reference#tag/upload/POST/v1/upload/analyze){target="\_blank"}.