> ## Documentation Index
> Fetch the complete documentation index at: https://algolia.com/llms.txt
> Use this file to discover all available pages before exploring further.

# NeuralSearch A/B testing

> Measure the impact of NeuralSearch by running controlled experiments before rolling out to all traffic.

Before you enable NeuralSearch for all search traffic,
run an A/B test to measure how it affects your search metrics.

NeuralSearch uses Algolia's standard A/B testing features.
You can create tests with the A/B testing API or from the Algolia dashboard.

## Choose a test type

You can run one of these test types:

| Test type               | Control        | Variant                           | When to use                                                                                                        |
| ----------------------- | -------------- | --------------------------------- | ------------------------------------------------------------------------------------------------------------------ |
| Keyword vs NeuralSearch | Keyword search | NeuralSearch                      | Compare keyword search with NeuralSearch                                                                           |
| Adaptive Intent test    | NeuralSearch   | NeuralSearch with Adaptive Intent | Compare NeuralSearch to NeuralSearch with [Adaptive Intent](/doc/guides/ai-relevance/neuralsearch/adaptive-intent) |

### Compare keyword search with NeuralSearch

Your index must have NeuralSearch enabled before you can launch an A/B test.
NeuralSearch can be in one of these states:

* **Inactive**: NeuralSearch isn't configured for the index.
* **Preview**: NeuralSearch is configured for testing but doesn't serve *all* search traffic.
* **Active**: NeuralSearch is enabled for the index's search traffic.

To change the NeuralSearch state for an index, set [`neuralSearchMode`](/doc/guides/ai-relevance/neuralsearch/get-started#parameters) appropriately.

This A/B test evaluates whether NeuralSearch improves your existing search experience.

* The control uses the index's default keyword search.
* The variant enables NeuralSearch.

```sh Command line expandable icon=square-terminal theme={"system"}
curl --request POST \
  --url https://analytics.algolia.com/2/abtests \
  --header 'content-type: application/json' \
  --header 'x-algolia-api-key: {ALGOLIA_ANALYTICS_API_KEY}' \
  --header 'x-algolia-application-id: {ALGOLIA_APPLICATION_ID}' \
  --data '{
    "name": "Keyword vs NeuralSearch",
    "variants": [
      {
        "index": "{INDEX_NAME}",
        "trafficPercentage": 50,
        "description": "Keyword search (control)"
      },
      {
        "index": "{INDEX_NAME}",
        "trafficPercentage": 50,
        "customSearchParameters": {
          "enableNeuralSearch": true
        },
        "description": "NeuralSearch (variant)"
      }
    ],
    "endAt": "2026-12-31T00:00:00.000Z"
  }'
```

### Compare NeuralSearch to NeuralSearch with Adaptive Intent

This A/B test measures whether [Adaptive Intent](/doc/guides/ai-relevance/neuralsearch/adaptive-intent) improves NeuralSearch results.

* The control uses NeuralSearch without Adaptive Intent.
* The variant enables Adaptive Intent.

```sh Command line expandable icon=square-terminal theme={"system"}
curl --request POST \
  --url https://analytics.algolia.com/2/abtests \
  --header 'content-type: application/json' \
  --header 'x-algolia-api-key: {ALGOLIA_ANALYTICS_API_KEY}' \
  --header 'x-algolia-application-id: {ALGOLIA_APPLICATION_ID}' \
  --data '{
    "name": "NeuralSearch vs NeuralSearch with Adaptive Intent",
    "variants": [
      {
        "index": "{INDEX_NAME}",
        "trafficPercentage": 50,
        "description": "NeuralSearch (control)"
      },
      {
        "index": "{INDEX_NAME}",
        "trafficPercentage": 50,
        "customSearchParameters": {
          "adiConfig": {
            "enabled": true
          }
        },
        "description": "NeuralSearch with Adaptive Intent (variant)"
      }
    ],
    "endAt": "2026-12-31T00:00:00.000Z"
  }'
```

## Start tests from the Algolia dashboard

To start NeuralSearch A/B tests from the Algolia dashboard:

1. Go to the [Algolia dashboard](https://dashboard.algolia.com/) and select your Algolia application.
2. Select your index.
3. Click the **NeuralSearch** tab.
4. Click **Launch an A/B test**.
5. Select the test type: **Keyword vs NeuralSearch** or **Adaptive Intent test**.
6. Set the traffic split and test duration.
7. Click **Create test**.

The Algolia dashboard pre-fills the required `customSearchParameters` for each test type, so you don't need to configure the API parameters manually.

## Test metrics

Algolia tracks these metrics automatically for every NeuralSearch A/B test:

* Conversion rate
* Click-through rate
* Add-to-cart rate
* No-results rate
* Average click position

For more information, see [Click and conversion events](/doc/guides/sending-events/concepts/event-types).

## Test duration

Run tests long enough to reach statistical significance:

* Tests should run for at least 14 days to account for weekly traffic patterns.
* Aim for a minimum increase of 3% on conversion rate.
* Indices with more traffic usually reach statistical significance faster.

## Stopping a test

Stop tests from the [Algolia dashboard](/doc/guides/ab-testing) or with the [A/B testing API](/doc/api-reference).
After you stop a test, traffic returns to the index's active configuration automatically.