How much answering the agent took off your hands
The number worth looking at first is how long the agent spent in conversations instead of you.
On a busy site that runs into thousands of hours. Divide it by a working week and you get a sense of the scale: this is time you did not spend typing the same answers.
It moves every day, and you can see the change over any period you pick.
Four numbers, updated as it works
There are four figures, each with the change over the period you selected:
- Chat sessions — how many conversations visitors started
- Messages received — how many messages they sent
- AI replies sent — how many answers the agent gave
- AI-handled chat time — how long those conversations took
Together they tell you whether the chatbot is being used, and how much work it is absorbing.
Any period you need
The four figures are shown for the period you pick rather than one fixed window, so you can look at a single busy day or at a long stretch.
Each number also shows the change over the period you selected, so you can see whether chats, messages and handled time are going up or down — not only what the total is.
What the numbers do not show
Being straight about the limits: Analytics counts activity. It does not break down what people asked about, and it does not attribute revenue.
For the content of conversations, go to Chat history — every conversation has a written summary, and the leads table shows what each contact was asking about.
For behaviour on the site around the chat, connect Google Analytics: it records chat opened, chat started, individual messages, and form started and submitted.
Tell us what to add
Analytics is deliberately small: four activity numbers, not a research suite.
If a figure is missing that would actually change how you run the widget, say so. The content of conversations and the leads table already live in chat history; this page is only the counts.