> For the complete documentation index, see [llms.txt](https://docs.codegiganten.de/plugin-documentation/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.codegiganten.de/plugin-documentation/en/apps/search-intelligent-and-ai-optimized/ai-features.md).

# AI features

Turn on, configure and understand hybrid search, AI preparation, synonym suggestions and insight analysis.

The **AI** area in the plugin dashboard bundles the functions that extend your search with meaning and with learning from real customer data. You open it in the sidebar in the group "Products & data". The page is called "AI features" and has the tabs "Settings", "Preview", "Synonym suggestions" (beta) and, once insight analysis is turned on, "Insight analysis".

Under "What does this tab control?" you find the short version: here you turn on and configure the AI functions of your shop search. Under "Preview" you test every setting before you activate it in the shop.

## Overview

| Function            | Purpose                                                             | Status |
| ------------------- | ------------------------------------------------------------------- | ------ |
| Hybrid search       | Adds a search by meaning to the word-exact search.                  | Beta   |
| AI preparation      | Prepares your products once for it. Prerequisite for hybrid search. |        |
| Synonym suggestions | The AI suggests synonyms that you accept with one click.            | Beta   |
| Insight analysis    | Evaluates clicks, cart, orders and search queries daily.            | Beta   |

In addition there are functions you set elsewhere: "Related hits" ("This might also fit"), substitute products and the rescue at zero hits (see [No results](https://docs.codegiganten.de/plugin-documentation/en/apps/search-intelligent-and-ai-optimized/no-results.md)) as well as the [Shopping assistant](https://docs.codegiganten.de/plugin-documentation/en/apps/search-intelligent-and-ai-optimized/shopping-assistant.md) (beta) with its own page.

## The tiles in the tab "Settings"

At the top, four tiles show the status at a glance:

* **"AI preparation":** Percent and "x of y products". While the value is unknown it says "being determined", in case of a disruption "disrupted" with the note "We are taking care of it."
* **"Hybrid search":** "Active" (with "Share of meaning …") or "Inactive".
* **"Open synonym suggestions":** Number of open suggestions and how many have been accepted so far.
* **"Hybrid search status":** "Available", "Limited", "Disrupted" or "Paused", each with a short explanation. When "Disrupted", the normal search takes over automatically.

## Hybrid search (beta)

The normal search compares words. Hybrid search also understands what is meant. The example from the plugin: whoever searches "cozy sofa" also finds the "couch with a soft cover". Both kinds of results are mixed.

### Turn it on, step by step

1. Wait until the "AI preparation" has reached 100 %. Before that, the switch is locked ("Can be turned on once the AI preparation has reached 100 %."). The progress bar updates every 10 seconds.
2. Turn on "Hybrid search" (beta). The default is off.
3. If needed, set the slider "Share of meaning" and click "Save". "Default" resets the slider, "View in preview" jumps to the tab "Preview".

The switch itself saves the last saved share along with it. Slider changes are only applied by "Save".

### Share of meaning

The slider ranges from "word-exact" to "meaning-based", from 10 % to 60 % in 5 % steps.

| Search mode | Default share of meaning            |
| ----------- | ----------------------------------- |
| Standard    | 30 %                                |
| Exploratory | 50 %                                |
| Precise     | none, it always searches word-exact |

If hybrid search is off, the slider is locked. You can still test the share in the preview.

### When the meaning search does not apply

In these cases the system simply continues with the normal search, and your customers see no error message:

* Inputs that look like article numbers.
* Sorting other than "Relevance".
* Search mode "Precise".
* Empty search.
* Very deep paging.
* When the meaning search is temporarily unavailable.

The beta notes in the plugin say: the function may be temporarily unavailable or only partly available. The normal search then steps in automatically. Your customers always get search results, so no errors arise in the search. In case of disruptions, the function can be paused temporarily, automatically or by us. The status then shows "Paused".

{% hint style="info" %}
The function is in the free beta phase and can be turned on or off at any time.
{% endhint %}

## Tab "Preview"

Under "How your search finds with AI" you compare what the meaning search changes:

1. Enter a search term and click "Compare".
2. On the left is "Text search only", on the right "Hybrid search (x % meaning)", 12 hits each.
3. The marker "new" shows hits that only occur on the right. The arrows ▲ and ▼ show the rank change.

The preview also works with hybrid search turned off and with slider values you have not saved yet. Try terms that do not appear literally in your product texts. That is where the difference shows most clearly. If the meaning search cannot be applied to a query, a warning appears and the right side shows the text search.

## AI preparation

For the meaning search, each product is prepared once from its data. This happens on our side, not in your shop. In the plugin you see the progress as "AI preparation" with percent and product count. It is done at 100 %.

* A fixed duration cannot be given. It depends on the number of your products.
* While the preparation runs, a note says results do not have full quality yet.
* If it fails, it says "AI preparation failed. It will be tried again at the next sync."

## Synonym suggestions (beta)

Instead of collecting synonyms yourself, you let suggestions be made. Under "Where do the suggestions come from?" the plugin explains two sources:

* **"From product data":** Products with the same meaning that are named differently, for example "sneaker" and "trainer".
* **"From search behavior":** Search queries with the same meaning after which customers click the same products. The last 30 days are evaluated.

The suggestions are determined anew every night. With "Determine now" you start the determination immediately, at most once per hour. The prerequisite is a finished AI preparation. The status shows for example "Determination running …" or "Last determination completed" with date and number of new suggestions.

### Working through suggestions

1. Open the tab "Synonym suggestions". The icon on the tab shows the number of open suggestions.
2. The list shows up to 100 of the most certain suggestions, each with the terms, the origin ("from product data" or "from search behavior") and "x % certain".
3. Untick individual terms if one does not fit. At least two terms must remain.
4. Click "Accept" or "Reject".

Accepted suggestions land in your [Synonyms](https://docs.codegiganten.de/plugin-documentation/en/apps/search-intelligent-and-ai-optimized/synonyms.md) and take effect in the search after a few minutes. If there is nothing to do, it says "No open suggestions."

## Insight analysis (beta)

Insight analysis learns from what happens in your shop. You turn it on in the tab "Settings". Turning it on starts a first evaluation immediately, after that it runs every night. With "Evaluate now" you start it by hand, at most once per hour.

Orders, clicks on search results, cart actions and search queries are evaluated. Cancelled and unpaid orders do not count. This produces:

* **Popularity per product (0 to 100):** It arises from clicks, cart and orders of the last 30 days. With equally good hits, popular products appear higher.
* **Search suggestions:** The suggestions while typing learn from real searches.
* **Bestsellers and trends:** Bestsellers of the last 7, 30 and 90 days and overall, plus trending search terms of the last 7 days compared with the 7 before. Useful for synonyms, landing pages and campaigns.

In the tab "Insight analysis" you see the processing status, the next automatic evaluation, tiles such as "Items sold (30 days)", "Revenue of items (30 days)", "Products with popularity" and "Trends (7 days)", and the list "Bestsellers". If there was no visitor data in the last 30 days, popularity stays unchanged.

## Quota and limits

* Hybrid search, synonym suggestions and insight analysis are in the free beta phase. There is no separate quota for them in the AI area.
* Share of meaning: at most 60 %.
* Hybrid search does not apply to article numbers, to sorting other than "Relevance", or in the mode "Precise".
* Determining synonym suggestions and running the insight evaluation by hand: at most once per hour.
* In general there is a quota for "search queries". Exceeding it triggers no automatic block, we get in touch if it is greatly exceeded.

## Read more

{% content-ref url="/pages/jaQ6nXm2Ug0rov8AO38O" %}
[Synonyms](https://docs.codegiganten.de/plugin-documentation/en/apps/search-intelligent-and-ai-optimized/synonyms.md)
{% endcontent-ref %}

{% content-ref url="/pages/s1zsa8FBn4f6VaJ5pRxJ" %}
[Products in the dashboard](https://docs.codegiganten.de/plugin-documentation/en/apps/search-intelligent-and-ai-optimized/products-in-the-dashboard.md)
{% endcontent-ref %}

{% content-ref url="/pages/ONADSPn6JKPHifT1E0iD" %}
[No results](https://docs.codegiganten.de/plugin-documentation/en/apps/search-intelligent-and-ai-optimized/no-results.md)
{% endcontent-ref %}

{% content-ref url="/pages/EKzI7kx2ZAjSIJVZz2DI" %}
[Shopping assistant (beta)](https://docs.codegiganten.de/plugin-documentation/en/apps/search-intelligent-and-ai-optimized/shopping-assistant.md)
{% endcontent-ref %}


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.

Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:

```
GET https://docs.codegiganten.de/plugin-documentation/en/apps/search-intelligent-and-ai-optimized/ai-features.md?ask=<question>&goal=<user_goal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.
