---
site_search: "GET /ask?q={query} (read-only search over extend.ai and docs.extend.ai; returns links and quoted passages, not generated answers)"
title: "Insurance Document Processing & OCR API | Extend"
description: "Document processing for insurance. Extract ACORD forms, loss runs, and declarations pages, split claim packets, and flag altered claim documents with Verify."
canonical: https://www.extend.ai/insurance
---

# Insurance Document Processing & OCR API | Extend

> Search this site and the docs: [`/ask?q=…`](/ask?q=webhook+retries) returns matching pages from extend.ai and docs.extend.ai, with quoted passages.

## Insurance: Process submissions and claims at intake

Extract ACORD forms and loss runs, split claim packets, and flag altered documents before a claim is paid.

[Try for free](https://dashboard.extend.ai/) · [Book demo](https://www.extend.ai/get-started)

## Trusted by leading AI teams

Comulate, First American, Axle, Ascend

## Benchmark: RealDoc-Bench

1,359 prompts across 581 real-world documents.

Metric: Document Q&A accuracy from parsed output.

| System | Score |
| --- | --- |
| **Extend Parse 2.0** | **95.7%** |
| Gemini 3.5 Flash | 89.0% |
| Azure DI | 88.8% |
| AWS Textract | 70.5% |

[See the full benchmark](https://www.extend.ai/resources/parse-2-and-realdocbench-launch.md)

## Underwrite and pay claims on data you can trust

- **[Verify] Flag Altered Claim Documents** — One edited amount on a repair estimate changes the payout. Verify detects AI edits, in-painting, and post-signature changes before a claim is paid.
- **Read Every ACORD Field** — ACORD 25, 27, and 125 forms pack insured, policy, and coverage data into dense grids. Extract named insureds, limits, and checkboxes into your schema, with a citation for every value.
- **Extract Loss Runs from Any Carrier** — Every carrier formats loss runs differently, with years of claims per report. Pull each claim's date of loss, status, and paid and reserved amounts into one consistent array.
- **Split Claim Packets Automatically** — Claim packets bundle FNOL forms, police reports, medical bills, photos, and repair estimates into one PDF. Split separates each document, even when one starts mid-page, so every page reaches the right extractor.
- **Compare Coverage Across Policies** — Declarations pages bury limits, deductibles, and endorsements in carrier-specific layouts. Extract policy numbers, effective dates, and limits into one schema for every carrier.

### Go deeper

**Benchmarks**

- [Verify: 96.3% of fraud flagged at a 4% false-positive rate](https://www.extend.ai/resources/introducing-extend-verify.md)
- [Document splitting benchmark](https://www.extend.ai/resources/document-splitting-benchmark.md)

**Guides**

- [ACORD form extraction guide](https://www.extend.ai/resources/acord-form-data-extraction-api.md)
- [Document splitting APIs for claims](https://www.extend.ai/resources/best-document-splitting-apis-insurance-legal-healthcare.md)

## How to get started

### 1. Create an account

Sign up at dashboard.extend.ai. New accounts get 10,000 free credits.

Copy an API key from the Developers page and export it as EXTEND_API_KEY.

[Sign up](https://dashboard.extend.ai/) · [Developers page](https://dashboard.extend.ai/developers)

### 2. Start from a template

Open the ACORD form classifier template and run it on your own files, then edit the schema to match your fields.

[ACORD form classifier](https://www.extend.ai/templates/acord-form-classifier) · [Browse all templates](https://www.extend.ai/templates)

### 3. Process your first document

Send a claim packet by URL and get back markdown for every page. Replace the URL with one of your files.

```bash
curl -X POST https://api.extend.ai/parse \
  -H "Authorization: Bearer $EXTEND_API_KEY" \
  -H "x-extend-api-version: 2026-02-09" \
  -H "Content-Type: application/json" \
  -d '{ "file": { "url": "https://your-bucket.s3.amazonaws.com/claim-packet.pdf" } }'
```

### 4. Move to production

Send production traffic to POST /parse_runs; the synchronous POST /parse times out after 5 minutes.

Turn on zero data retention at no extra cost, and run in Extend's US or EU cloud or in your own VPC.

[API docs](https://docs.extend.ai/)

## Popular templates

- **ACORD form classifier**: Classifies ACORD forms, then extracts applicant details, coverage selections, and policy terms. [Use template](https://www.extend.ai/templates/acord-form-classifier)
- **Loss run report**: Extracts claim details, injury descriptions, and financial summaries from loss runs. [Use template](https://www.extend.ai/templates/loss-run-report)
- **Certificate of liability insurance**: Extracts the insured, insurers, coverage, and certificate details from COIs. [Use template](https://www.extend.ai/templates/certificate-of-liability-insurance)

[Browse all templates](https://www.extend.ai/templates)

## Frequently asked questions

### Which insurance documents does Extend handle?

ACORD 25, 27, and 125 forms, loss runs, declarations pages, certificates of insurance, FNOL forms, and claim packets with police reports, estimates, and medical bills. Teams like Comulate, First American, Axle, and Ascend use Extend. [Read the ACORD extraction guide](https://www.extend.ai/resources/acord-form-data-extraction-api.md).

### Can it catch altered claim documents?

Yes. Verify checks each file's pixels, structure, and revision history for AI-generated documents, AI edits, patched text, and changes made after signing. On Extend's internal benchmark it flagged 96.3% of fraudulent documents at a 4% false-positive rate. It does not verify identity. [Read the Verify launch post](https://www.extend.ai/resources/introducing-extend-verify.md).

### How does it handle claim packets?

Splitting separates a packet into its claim forms, police reports, estimates, and medical bills, even when one starts mid-page, and holds accuracy on files of 2,000+ pages.

### Can it extract every claim from a long loss run?

Yes. Array extraction returns every claim with its date of loss, status, and paid and reserved amounts. On LongArray-Extract, a public benchmark of 45 PDFs with up to about 2,200 rows each, Extend's Operator-1 scores 99.3%, the top result. [See the LongArray-Extract results](https://www.extend.ai/resources/long-array-extraction-benchmark.md).

### Is Extend secure enough for policyholder data?

Yes. Extend is SOC 2, HIPAA, and GDPR compliant, with regular third-party penetration tests. Zero data retention is free, and you can run in Extend's US or EU cloud or in your own VPC.
