Captcha OCR

Captchas and accessibility

2026-10-03 · Captcha basics

A picture-only test excludes people who cannot see it. Audio and simpler questions were added for that reason.

The first captchas assumed a person who can see and type. Distorted letters fail that assumption for a blind visitor, and a faint low-contrast image fails it for many other people. A test that a portion of the public cannot complete is a blocked door, not a security check.

Audio captchas were the usual repair. The site speaks a short code through noise, and the visitor types it. That helps someone who can hear and type, and it fails someone who is deaf and blind, or who is in a place where they cannot play sound. Audio captchas describes the clip. This API does not transcribe it.

Other sites ask a plain question, such as a small sum written as real text in the page, or they offer a checkbox with an accessible name. A sum that is already text in the HTML does not need an image reader. Read the DOM. Call an OCR API only when the challenge is painted into a picture.

If you run a form, put the captcha on the route that is abused, offer a path a screen reader can use, and keep the image's text alternative honest. An empty alt on the only challenge means the accessible name of the page is a blank.

Captcha OCR is a reader for images you already have. It is not an accessibility layer for a checkbox vendor. For the images it does read, the types page is the map.

Ready to call it? Start with the quickstart, the API guide, or pricing.

More in Captcha basics