> For the complete documentation index, see [llms.txt](https://documentation.ajaxsearchpro.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://documentation.ajaxsearchpro.com/search-ai.md).

# Search AI

### What is Search AI? <a href="#what-is-search-ai" id="what-is-search-ai"></a>

**Search AI** is a suite of AI-powered tools built into Ajax Search Pro that help you *configure*, *diagnose* and *improve* your search without guessing. Instead of digging through settings and statistics by hand, you can let the assistant analyse your content, your index and your real search history, then tell you what's wrong and what to change.

The suite includes tools that:

* **set up a new search instance for you** from a few simple questions (Setup Wizard),
* **explain why a specific search returns nothing** (Diagnostics),
* **report on the health of your search** over the last weeks (Quality Report),
* **find configuration problems** in your priority rules and stopwords,
* and **generate keyword suggestions and synonyms** grounded in your own site.

You'll find everything under **Ajax Search Pro → Search AI** in the WordPress admin menu.

<figure><img src="/files/9AJqjyYojjIQolnafhAG" alt=""><figcaption></figcaption></figure>

#### Requirements & how billing works <a href="#requirements--how-billing-works" id="requirements--how-billing-works"></a>

* **An active license is required.** AI features cannot be enabled until your license key is activated. If it isn't, the panels show a prompt to activate it first.
* **AI usage is metered in tokens.** Every plan includes a **monthly token budget**; each AI action consumes some of it. You can watch your consumption on the [Usage](#usage) tab. Heavier operations (like generating suggestions for many instances at once) warn you before spending a large share of the budget.
* **Results are cached.** Reports, synonym suggestions and wizard configurations are cached, so re-opening a recent result doesn't cost more tokens. You can force fresh results by purging the AI cache.

***

### Settings <a href="#settings" id="settings"></a>

The **Settings** tab is where you turn AI features on and off. Each feature has its own toggle, plus an **Enable all features** switch at the top to turn the whole suite on or off at once. After changing toggles, click **Save**.

The available features are:

| Feature                        | What it does                                                                                                |
| ------------------------------ | ----------------------------------------------------------------------------------------------------------- |
| **Search Diagnostics**         | Conversational "Why isn't X showing up?" assistant. Checks stopwords, index, exclusions and priority rules. |
| **AI Synonym Discovery**       | Suggests creative synonyms — alternative words real users might type for the same content.                  |
| **Search Quality Report**      | Analyses the last 30 days of search statistics and produces an actionable performance report.               |
| **AI Setup Wizard**            | Guides you through creating a new search instance based on your site type and requirements.                 |
| **Priorities Health Check**    | Scans priority rules for orphaned term/post references and suggests new rules based on popular searches.    |
| **Stopword Conflict Detector** | Cross-references zero-result searches with the active stopword list to flag silently dropped keywords.      |
| **AI Keyword Suggestions**     | Pre-generates grounded autocomplete and keyword suggestions from your site's index and search history.      |

<figure><img src="/files/0idCKGtkLYskic0vnviR" alt=""><figcaption></figcaption></figure>

#### AI cache <a href="#ai-cache" id="ai-cache"></a>

Below the toggles, the **AI Cache** section has a **Purge Cache** button. This clears all cached AI responses (quality reports, synonym suggestions, wizard configs) so they're regenerated fresh on next use. Use it when your content or settings have changed significantly and you want the AI to look at the current state rather than a cached snapshot.

<div><figure><img src="/files/yJM2dQuvFYu0kkJq1qwu" alt="" width="375"><figcaption></figcaption></figure> <figure><img src="/files/ekcKbJThldaj8B67DE5g" alt="" width="375"><figcaption></figcaption></figure></div>

***

### Setup Wizard <a href="#setup-wizard" id="setup-wizard"></a>

The **Setup Wizard** builds a complete, ready-to-use search instance for you. It walks through a few short steps and then asks the AI to generate a tailored configuration.

The steps are:

1. **Site type** — WooCommerce store, Blog / News, Documentation, Corporate, or Mixed. This tells the AI what kind of content to optimise for.
2. **Content** — which content types the search should cover (posts, pages, products), and whether to include user profiles and attachments/PDFs.
3. **Search engine** — choose the **Standard** (real-time MySQL) engine or the **Index Table** engine. When you pick Index Table, you can also configure taxonomy-term search, which custom fields to index, and PDF/file content search.
4. **Behavior** — keyword matching (all words / any word), whether to search within post content, whether to show descriptions in results, and the primary language.
5. **Review** — the AI returns a suggested configuration with a plain-language summary and a list of **key decisions** explaining *why* it chose each setting. Give the instance a name and create it.

Once created, you get a link to open the instance list and fine-tune it. If the wizard changed the **Index Table** configuration, it shows a **re-index reminder** with a shortcut to the Index Table page — run a new index so the changes take effect.

<div><figure><img src="/files/C7MEY1FyYCjleWikm1TI" alt="" width="330"><figcaption></figcaption></figure> <figure><img src="/files/BOqgnNgxZY1svFUaLgYN" alt="" width="325"><figcaption></figcaption></figure></div>

***

### AI Keyword Suggestions <a href="#ai-keyword-suggestions" id="ai-keyword-suggestions"></a>

This tool **pre-generates keyword suggestions** for your search instances, grounded in your own index and search history (not generic web data). The generated words can then power your autocomplete and "did you mean" keyword suggestions.

How to use it:

1. Select one or more **instances** (leave empty to generate for all).
2. Set **Max suggestions** per instance (10–1000).
3. Click **Generate**. Generating for many instances or with a high suggestion count can use a significant share of your monthly tokens — the tool warns you with an estimate before proceeding.

Each instance's generated words appear as removable chips, with the count and generation date. You can remove individual words or **Clear all** for an instance.

{% hint style="warning" %}
**Important:** generating the words is only half the job. To actually use them, add **AI Suggestions** as a source in each instance's **Autocomplete** or **Keyword Suggestions** settings. On the frontend, AI-sourced phrases are flagged with a sparkle badge.
{% endhint %}

<figure><img src="/files/XPkKjcWgLOTCF70PiaYn" alt=""><figcaption></figcaption></figure>

***

### Priorities Advisor <a href="#priorities-advisor" id="priorities-advisor"></a>

The **Priorities Advisor** (Priorities Health Check) scans a search instance's priority rules and flags configuration problems, then offers AI recommendations — including new rules suggested from your popular searches.

Pick an instance and click **Run Analysis**. It reports issues such as:

* **Orphaned taxonomy terms** — rules pointing at terms that no longer exist.
* **Orphaned post references** — rules pointing at deleted posts.
* **Empty rules** — rules that target nothing.
* **Misconfigured phrase triggers.**
* **Missing group names.**

Results appear as health chips (or a green "No issues detected") followed by the AI's written analysis and recommendations.

<figure><img src="/files/ZzUjRpRZSx7YIGnwG3Fp" alt=""><figcaption></figcaption></figure>

***

### Stopword Detector <a href="#stopword-detector" id="stopword-detector"></a>

The **Stopword Conflict Detector** finds cases where your **stopword list is silently dropping words that visitors actually search for**. It cross-references zero-result searches from the selected time period against the active stopword list.

Pick an instance, choose the period, and click **Run Analysis**. For each conflicting stopword it shows:

* a **risk badge** (low / medium / high) indicating how safe it is to remove,
* the number of **affected searches**,
* an AI **explanation** and **recommendation**,
* the list of affected zero-result phrases.

You can **Remove** a conflicting word from the stopword list right from the card. After removing words, a re-index is recommended so the change applies to already-indexed content.

{% hint style="info" %}
**Note:** this feature requires stopword filtering to be enabled in the Index Table settings. If it's off, the panel tells you so.
{% endhint %}

<figure><img src="/files/QptQ4mrlyXfpEYANb9rB" alt=""><figcaption></figcaption></figure>

***

### Quality Report <a href="#quality-report" id="quality-report"></a>

The **Search Quality Report** analyses the last 30 days (configurable) of search statistics for an instance and produces an actionable performance report.

Pick an instance, set the number of days, and click **Generate Report**. You get:

* **Key metrics** — total searches, unique phrases, zero-results rate, click-through rate, and average results returned (colour-coded green/amber/red so problems stand out).
* **Breakdown by device** — desktop, mobile and tablet volumes with their zero-result rates.
* **An AI-written report** with concrete recommendations.
* **Top zero-results phrases** — the searches that most often returned nothing. Each row has a **Discover** button (to find synonyms for that phrase) and a **Diagnose** button (to jump straight into Diagnostics for it).

<figure><img src="/files/zecy4VL5OJHCxfdacGsK" alt=""><figcaption></figcaption></figure>

***

### Diagnostics <a href="#diagnostics" id="diagnostics"></a>

**Search Diagnostics** is a conversational assistant that answers the question *"Why isn't this showing up?"* for a specific search.

Enter the **search phrase** a visitor typed and select the **search instance**, then click **Run Diagnostic**. The AI inspects your actual configuration — index hits, tokenization (including words dropped as stopwords or for being too short), keyword logic, taxonomy/date/user/custom-field filters, exclusion rules and matching priority groups — and explains in plain language why results may be missing or unexpected.

* A collapsible **Diagnostic context** panel shows the raw facts the AI based its answer on.
* You can ask **follow-up questions** in a chat, so it's an interactive investigation rather than a one-shot answer.
* Recent diagnostic sessions are kept in a **History** sidebar so you can revisit them.

<figure><img src="/files/fxKFQBer3zautPxXgtQQ" alt=""><figcaption></figcaption></figure>

***

### AI Synonym Discovery <a href="#ai-synonym-discovery" id="ai-synonym-discovery"></a>

**AI Synonym Discovery** suggests alternative words real users might type for the same content, so you can add them as synonyms and catch more searches. Unlike the other tools, it doesn't have its own tab — instead, a **Discover** button (sparkle icon) appears wherever a search phrase is shown and synonyms would help:

* on **zero-result phrases** in the [Quality Report](#quality-report),
* on **popular searches** in the Statistics panels,
* and directly in the **Synonyms** management page.

Click **Discover** next to a phrase to open a popover of suggested synonyms, each with the reason it was suggested and how many times it occurs in your index. Click **Add** to apply a synonym instantly.

{% hint style="info" %}
**Note:** this requires the **AI Synonym Discovery** feature to be enabled in [Settings](#settings). If it's off, the popover links you there.
{% endhint %}

<figure><img src="/files/Td9mfeFRygLFj0Xr0sKg" alt=""><figcaption></figcaption></figure>

***

### Usage <a href="#usage" id="usage"></a>

The **Usage** tab shows your **monthly token consumption**:

* tokens used this month versus your plan's cap, with a progress bar that turns amber and then red as you approach the limit (plans without a cap show "Unlimited usage"),
* your **product/plan name**,
* a **Usage by feature** breakdown (from the local interaction log, last 30 days) showing which features consumed the most tokens and how many calls each made.

Use **Refresh** to pull the latest figures.

<figure><img src="/files/Tv90CnO0SDBhGoIvNztr" alt=""><figcaption></figcaption></figure>

***

### Interaction Log <a href="#interaction-log" id="interaction-log"></a>

The **Interaction Log** records every AI call made from your site — the feature used, a short prompt summary, tokens consumed, whether the result was served from cache, and when it happened. You can **filter by feature** and page through the history.

It's useful for auditing what the AI has been doing and where your tokens are going.

<figure><img src="/files/e6kkJlQcJEGrP2CdjtMK" alt=""><figcaption></figcaption></figure>

***

### Tips & notes <a href="#tips--notes" id="tips--notes"></a>

* **Enable only what you need.** Each feature is independent — turn on just the tools you use to keep things simple.
* **Re-index after stopword changes.** Removing a stopword (via the Stopword Detector) only fully applies once you rebuild the [Index Table](/index-table.md).
* **Generated suggestions need to be wired up.** AI Keyword Suggestions and discovered synonyms don't affect the frontend until you add **AI Suggestions** as a source (for keywords) or save the synonyms.
* **Purge the cache after big changes.** If you've significantly changed your content or configuration, purge the AI cache so reports reflect the current state.


---

# 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 dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://documentation.ajaxsearchpro.com/search-ai.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
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.