# Amazon Textract Insights launches with AI-powered document extraction

> Amazon adds a generative‑AI feature that lets users describe fields in plain language and get structured data with confidence scores and source grounding.

Oossa · 2026-11-20 · https://oossa.com/en/amazon-textract-insights-launches-with-ai-powered-document-extraction

Amazon announced the general availability of Amazon Textract Insights, a new API that adds generative AI to its document‑processing service. Users can write natural‑language descriptions of the data they need, submit a single PDF or image file, and receive a JSON result where each value includes a confidence score and page‑level grounding. The feature works on mixed packets up to 3,000 pages and requires no model training. Textract Insights is priced per page and is currently offered in the same AWS regions as the core service, except the two new regions added for Textract Core.

The service also supports pre‑built blueprints for common forms like W‑2s and passports, or custom blueprints for proprietary documents. Results are written back to the user’s S3 bucket, encrypted at rest, and can be reviewed only for fields below a chosen confidence threshold.

## The facts

- General availability announced on 2026-11-20
- Handles documents up to 3,000 pages per request

## Why it matters

Financial and health‑care teams can automate more document checks while still auditing low‑confidence fields, reducing manual review workload.

## Sources & references

1. [Announcing Amazon Textract Insights: Verifiable generative AI for documents](https://aws.amazon.com/blogs/machine-learning/announcing-amazon-textract-insights-verifiable-generative-ai-for-documents/) – AWS, 2026-11-20

Last updated: 2026-11-20
