The AI Concierge pushes events to your Google Tag Manager data layer whenever a guest opens the chat or gets results, so you can measure assistant usage in your existing analytics setup.
The AI Concierge announces its activity to your analytics so you can measure how guests use it. Every time the chat opens and every time it produces a set of results, it pushes an event to the browser data layer for Google Tag Manager, and dispatches a matching browser event for any tooling that listens that way. Both are passive and never interfere with the chat.
The events #
| Data layer event | Browser event | Fires when |
|---|---|---|
| hr_ai_concierge_opened | hr-ai-concierge:opened | The chat opens (one per open). |
| hr_ai_concierge_results | hr-ai-concierge:results | The assistant produces a set of results (every search, any count). |
Every payload includes a per-conversation id, hr_ai_session_id, so you can group events into sessions. The results event carries extra detail:
| Field | Meaning |
|---|---|
| result_count | How many properties matched. |
| on_explorer | True when results were applied in place on the search page. |
| redirected | True when the response sent the guest to the results page. |
| handpicked | True when the assistant curated specific listings rather than only setting filters. |
Counting results pages in GTM #
- Create a new trigger of type Custom Event with the event name
hr_ai_concierge_results. - Add a tag (a GA4 event, for example) that fires on that trigger. Map
result_count,handpicked, andhr_ai_session_idas parameters if you want them in your reports. - Open GTM Preview, load your site, open the assistant, and run a search. You will see the event arrive in the data layer, and your tag fire.
Use hr_ai_concierge_opened the same way to count how many times the assistant is opened. To count unique conversations rather than raw opens, group by hr_ai_session_id.
A note on the older marker #
You may still occasionally see a temporary hr_ai_ask parameter appear on the page URL. That happens when a guest submits a question through a fallback path that reloads the page instead of updating the chat in place, most often when the ask bar’s fallback submit target is left blank and the in-page assistant is not present on that page. It is not an analytics event and is not meant to be tracked; the data layer events described here are the stable, intended way to measure the assistant, and they work whether it filters results in place or sends the guest to a results page.
Listening without Tag Manager #
If you do not use GTM, the same information is available as standard browser events. Listen for hr-ai-concierge:opened and hr-ai-concierge:results on the window object; the event detail carries the same payload described above.
Tips #
- Report on
hr_ai_concierge_resultsfirst; it is the event most tied to bookings, since it fires every time the assistant actually surfaces properties. - Group both events by
hr_ai_session_idif you want conversation-level counts instead of raw event counts. - See AI Concierge: Performance for the built-in reporting the plugin already gives you, without any GTM setup.
- New to the AI Concierge? Start with the AI Concierge overview.