Types of captcha explained, and which ones a solver can handle
There are several main types of captcha. Some ask the visitor to read or pick something, some run checks in the background and some combine a widget with a server-side check. The type decides how a script can get past it, and which type a solver supports decides whether you can use one. This guide lists the main types, explains how each one works, and states the current support status for Scarnox.
The short version: today Scarnox has one live task type, PopularCaptcha, proxyless. Other types are coming soon, and the sections below say which.
Support status at a glance
| Captcha type | How it works | Scarnox status |
|---|---|---|
| PopularCaptcha widget | Widget that writes a token to h-captcha-response | Live (proxyless) |
| Text and image challenges | Read or pick characters and objects | Not a live task type |
| Audio challenges | Transcribe a spoken clip | Not a live task type |
| Google reCAPTCHA v2 | Checkbox, with image challenges when flagged | Coming soon |
| Google reCAPTCHA v3 | Score from behaviour, no challenge | Coming soon |
| Cloudflare Turnstile | Non-interactive browser checks | Coming soon |
| Arkose Labs (FunCaptcha) | Game-style enforcement challenges | Coming soon |
| GeeTest | Slider and shape puzzles | Coming soon |
| AWS WAF captcha | Puzzle challenges from AWS WAF | Coming soon |
Check the captcha solver API guide and the API reference before you build against any type. This table reflects the status at the time of writing.
Text captchas
A text captcha shows distorted characters and asks you to type them. The distortion is meant to defeat simple optical character recognition. These are the oldest type and are now rare on major sites, but you still find them on smaller forms and legacy systems.
Image captchas
An image captcha asks you to pick every tile that contains an object, such as traffic lights, crosswalks or storefronts. The challenge is a classification task, and a sophisticated version may ask for several rounds. Scraping and automation tools often hit these on high-traffic sites. The image and text captcha OCR guide covers the reading side of this problem.
Audio captchas
An audio captcha plays a spoken sequence of digits or words, usually as an accessibility option. The sound is distorted with noise. These exist mainly as an alternative for users who cannot see the visual challenge, which is one reason they should be offered alongside accessible alternatives rather than automated away.
Puzzle captchas
Puzzle captchas ask for a physical-style action: slide a piece into a gap, rotate an image to upright, or drag a shape. Slider puzzles are common, and GeeTest is a well-known provider of them. The answer comes from a position or an angle, and the challenge often checks how the drag was made, not only where it ended.
Invisible and score-based captchas
Some captchas run without a visible challenge. The browser performs checks in the background, and the site decides from the result whether to show a challenge at all. This is what people mean by "invisible captcha". reCAPTCHA v3 goes further and returns a score rather than a pass or fail. The invisible captcha guide explains how the invisible mode affects a solver integration, and the reCAPTCHA v3 score guide covers scoring in detail.
Token widgets
A token widget is a widget in the page that completes its checks and writes a token into a hidden form field. The site sends that token to the provider, or to its own backend, for verification. PopularCaptcha, reCAPTCHA v2 and Cloudflare Turnstile all use this pattern.
This is the type a solver can handle without touching the visual challenge. The solver produces the token, and your code writes it into the form. That is why the PopularCaptcha integration is only a few lines long. The solve captcha with Python guide shows the flow.
Enforcement challenges
Some large sites use enforcement challenges that combine a risk score with a game-like puzzle. Arkose Labs, which is often called FunCaptcha, is the best-known example. The puzzles change often and the risk engine is tuned per site, which makes them hard to handle with a generic approach.
Why the type matters for your code
Each type needs a different kind of solver:
- Content challenges need a model that can read or classify what is on the screen.
- Puzzle challenges need a model that estimates a position or angle, plus a realistic drag.
- Token widgets need a solver that produces a valid response for the provider.
A solver that claims to handle "all captchas" is worth questioning. Ask which types it supports live, and what it does with the ones it does not support.
What to do with an unsupported type
If the captcha you need is not on the live list, you have three realistic options:
- Reduce how often you trigger it, using the approaches in the avoid captchas when scraping guide.
- Use a test key or a staging environment if the captcha guards your own system.
- Ask the site owner for an API, an allowlist or an accessible alternative.
Support for the coming-soon types is in progress. Create an account to start with free trial credit, no card needed, and you will be ready to use each type when it ships.