AI site search
A search box embedded in your website: visitors search your own content and, when they ask a question, get a short answer with sources from the pages found. Available on Growth and above…
A search box embedded in your website: visitors search your own content and, when they ask a question, get a short answer with sources from the pages found. Available on Growth and above; AI answers need the AI assistant to be enabled for the organization (ADR-0093).
How it works
- Your pages are indexed: your sitemap is crawled, or you upload documents through the API.
- While the visitor types, the search runs on a Postgres full-text index (by word stems; no AI, always available).
- When the visitor asks a question (Enter or "Ask AI"), a short answer is written from the best matching sections. The answer relies only on your site's content and shows the source pages; if the content has no answer, it says so.
- If an AI answer cannot be given (off, quota or cap reached), the search results are shown as usual; the visitor sees no error.
Setup
Dashboard → Web growth tools → Site search.
- In the Sources tab, add your sitemap (
https://www.example.com/sitemap.xml) or a list of addresses. The first crawl finishes within a few minutes. - In the Settings tab, turn site search on; turn AI answers on too if you want them.
- In the Try it tab, check the results and answer a visitor would see.
- Add it to your site:
<button id="site-search-button">Search</button>
<script>
Pushizo.Search.attach('#site-search-button', { hotkey: true });
</script>| Setting | Description |
|---|---|
| Site search | While off, the search endpoint does not answer (403). |
| AI answers | An answer with sources when a question is asked. While off, only the result list. |
| Content language | The language used for word-stem matching. Changing it recomputes the index; no re-crawl is needed. |
| Monthly token cap | The most tokens AI answers can spend in this app per month (default 500,000). |
| Allowed origins | Sites that may call search. If empty, the app's allowed origins apply. |
| Query retention | Days query logs are kept for the report (7–365, default 90). |
Sources and crawling
- Sitemap:
sitemap.xmlor a sitemap index. Only pages on the same site are crawled. - Address list: up to 500 addresses.
- Only public
httpsaddresses are crawled; private network addresses are rejected. Redirects are followed only within the same site. - The crawler identifies itself as
PushizoBotand followsrobots.txtrules and thenoindextag:
User-agent: PushizoBot
Disallow: /account/
Disallow: /cart- Menus, headers and footers are not indexed; if the page has
<main>or<article>, only that part is read. - Pages are re-crawled at the interval you set (1–720 hours). Unchanged pages are not rewritten, pages no longer on the site are removed from the index, and the previous version of a page with a temporary error is kept.
- Limits: up to 2000 pages per source, 4 MB per page, 10,000 documents per app.
| Status | Meaning |
|---|---|
| Queued / Crawling | The crawl is waiting or running. |
| Up to date | All pages are indexed. |
| Partly crawled | Some pages could not be fetched; the error code is shown on the source (e.g. http_404, timeout). |
| Failed | The sitemap could not be read or no page could be fetched (e.g. sitemap_http_500, robots_unreachable). |
Uploading documents through the API
Upload content that cannot be crawled (help articles behind a login, product data) through the API. The key is unique within the app; uploading again with the same key replaces the document.
curl -X PUT "https://api.pushizo.com/v1/apps/$APP_ID/site-search/documents/shipping" \
-H "Authorization: Bearer $PUSHIZO_API_KEY" -H "Content-Type: application/json" \
-d '{
"url": "https://www.example.com/shipping",
"title": "Shipping and delivery",
"content": "# Shipping fee\nShipping is free for orders over 50 USD.\n# Delivery time\nOrders ship within 2 business days."
}'Provide either content (plain text; lines starting with #, ##, ### are section headings) or html. To delete,
use DELETE /v1/apps/{app}/site-search/documents/{key}.
Search window and headless use
The ready-made window is a separate bundle loaded on first open; it is keyboard accessible (arrow keys, Enter, Esc) and supports right-to-left languages and 22 languages.
// Ready-made window
await Pushizo.Search.open({ placeholder: 'Products, shipping, returns…', theme: 'auto', color: '#2563eb' });
// Your own UI
const res = await Pushizo.Search.query('how much is shipping', { mode: 'answer', limit: 5 });
res.results; // [{ id, url, title, heading, snippet, snippet_html }]
res.answer; // { text, sources: [{ id, url, title }] } or null
res.answer_status; // 'answered' | 'not_found' | null
await Pushizo.Search.trackClick(res.query_id, res.results[0].url, 1);Details: Web SDK → Site search.
Cost and security
- Rate limit: 120 searches and 10 AI answers per minute per IP; 12,000 searches per minute per app.
- Token cap: once the app's monthly cap is reached, only search results are returned. The organization's AI quota and plan credits also apply.
- Origin list: if a list is set, only requests from those sites are answered.
- Answer limits: the model sees only the sections found; another tenant's content never enters the prompt. The answer passes the banned-phrase list; if it matches, it is not shown.
- Typing only searches; an AI answer is generated only when the visitor asks for it.
Report and personal data
The Report tab shows the last 30 days: number of searches, searches with no results, click-through rate, AI answer counts, top searches and queries with no results. Queries with no results are the fastest way to find missing content.
- Query text is masked before it is saved and before it reaches the model: emails, phone numbers, IBANs and numbers longer than 6 digits are replaced with placeholders.
- No visitor identity, subscription or IP address is stored; query logs cannot be linked to a person.
- Logs are deleted when the retention period ends; "Clear data" also deletes query logs.
API summary
| Endpoint | Description |
|---|---|
GET /v1/apps/{app}/site-search | Settings, index status, AI usage |
PATCH /v1/apps/{app}/site-search/settings | Update settings |
GET/POST /v1/apps/{app}/site-search/sources | List / add sources |
PATCH/DELETE /v1/apps/{app}/site-search/sources/{id} | Edit / delete a source |
POST /v1/apps/{app}/site-search/sources/{id}/crawl | Re-crawl now |
GET /v1/apps/{app}/site-search/documents | Indexed documents |
PUT/DELETE /v1/apps/{app}/site-search/documents/{key} | Upload / delete a document |
GET /v1/apps/{app}/site-search/report | Search report |
POST /sdk/v1/apps/{app}/search | Visitor search (keyless; used by the Web SDK) |
Read endpoints need the analytics:read scope, management endpoints admin.
Custom AI agents
Agents that follow your own instructions, review segments and campaign results with the tools you allow and prepare drafts for your approval. They run on a schedule (cron) or manually…
Journeys (Phase 11)
A journey is an automated flow that enrolls a user through a trigger and moves them through wait, condition, branching and send steps (the equivalent of the legacy provider's Journeys). In the API, the resource name is journeys, the field name is journey_id, and the event…