---
description: "URL split tests send visitors to two or more different URLs, such as a new landing page against the current one, and compare them like any A/B test. How to set one up, how the redirect works, and what search engines see."
---

# URL split tests

::: info In short
- **What:** compares whole pages at different URLs, such as a redesigned landing page against the current one.
- **How:** visitors to the original URL are split evenly; some stay, the others are redirected to a variant URL.
- **Results:** read like any test, with the same planned window and winner.
- **When:** the variant is a separate page you built. For changes to one page, use a normal test.
:::

## Set one up

1. Build the variant page on your own site, with the Dalton script on it.
2. In Dalton, click **New experiment ▾ → URL split test**.
3. Give it a name, enter the **Original** URL and one or more **Variant** URLs.
4. **Create draft**, then launch it like any other test.

The test then sits on the Experiments page next to your other tests, with the same [results](/experiments/reading-results) and [planned window](/experiments/how-experiments-run).

## How the redirect works

- Only visitors who land on the **exact original URL** enter the test. Other pages, and other paths that start the same way, aren't affected.
- Visitors in a variant are redirected in the browser to the variant URL. **Query parameters are kept**, so UTM tags and ad click IDs still reach the variant page.
- While the redirect happens, the original page stays hidden, so visitors don't see it flash.
- A visitor keeps their URL for the visit (24 hours). Coming back to the original URL sends them to the same variant, and the back button doesn't loop them.

## Search engines

Dalton doesn't run for search engine crawlers, so they see the original URL and are never redirected. Give each variant page a `rel="canonical"` link to the original URL, or set it to `noindex`, so it isn't indexed as a duplicate. This follows [Google's guidance on website testing](https://developers.google.com/search/docs/crawling-indexing/website-testing).

## When to use which

| You want to test | Use |
| --- | --- |
| Copy, layout, images or sections on one page | A normal test in the [editor](/editor/using-the-editor) |
| The same change across a template | A normal test with a wide [targeting group](/experiments/targeting#testing-across-many-pages) |
| A completely different page you built yourself | A URL split test |