Website Chatbots
How the chatbot works out which tribe a visitor means
Where the chatbot gets a tribe from, what happens when a visitor names a whole city, and why it sometimes asks a question it looks like it should already know.
Last updated September 16, 2026
Almost everything the chatbot says depends on knowing which of your tribes it is talking about — a tribe is one location, branch or business unit inside your organisation. It decides which opening hours and phone number to quote, and — the part that matters most — which tribe receives the enquiry and the notification email at the end of the conversation. This guide explains where that answer comes from, so that when it gets one wrong you know which setting to look at.
Four ways it learns a tribe
The chatbot collects the tribe from whichever of these is available, and a stronger source always replaces a weaker one:
| Source | How it is treated |
|---|---|
| Remembered from earlier in the visit | Weakest. A guess, used only when nothing else is known, and confirmed in the conversation before any enquiry is submitted. |
| The page the visitor is on | The chatbot reads the tribe from the address of the page it is sitting on. This replaces a remembered tribe. |
| The embed snippet | If you have pinned a tribe on the snippet you pasted into the page, that is used instead of reading the address. |
| The conversation itself | Strongest. The visitor chose a tribe, corrected the chatbot, or an enquiry was submitted. Nothing later overrides this. |
When the visitor names a place
On a general page there is no tribe to read, so the chatbot asks the visitor where they are. What they type is looked up on the map, anchored to the country your business operates in — so "Margate" means Margate in South Africa for a South African network, and never Margate in England. The resulting point then goes through the same three-step search your website's store locator uses, described in How a customer is matched to a tribe: drawn territory first, then territory postcodes, then distance.
If the name is genuinely ambiguous — a town name used in several parts of the country, with candidates a long way apart — the chatbot asks which one the visitor means instead of guessing.
Naming a whole city
A city is a problem that a suburb is not. However large a place is, looking it up produces one point, near the middle, so a visitor who says "Pretoria" or "Sydney" is really pointing at the city centre. In a network where several tribes share one city, that means every enquiry from that city would go to whichever tribe covers the middle of it.
So the chatbot stops guessing in exactly that situation. When a visitor names a city or a region, and more than one of your tribes sits within about 30 kilometres of it, it offers up to three nearby tribes as cards the visitor can tap, each showing the tribe's name, its suburb and how far away it is. Tapping one ties the conversation to that tribe, and the conversation carries on from there. Distances are shown in the visitor's own units, so a visitor in a country that measures in miles reads miles and everybody else reads kilometres.
Anything narrower than a city is answered with a single tribe as before. A suburb, a town, a postcode and a street address are all specific enough to answer directly, and turning a precise address into a menu would be a step backwards.
Testing it
The test console on the chatbot's settings screen lets you hold a conversation as a visitor and choose which tribe to test as, including testing with no tribe at all. Testing with no tribe is how you see the questions a visitor on a general page is actually asked, and how you check that a city name offers the tribes you expect.
If your tribes have no coordinates
The search needs a latitude and longitude on each tribe. Where an organisation has none, no place a visitor names could ever match, and asking again would loop forever — so the chatbot falls back to a single default tribe and carries on, which at least lets the enquiry be captured. If the chatbot keeps attributing enquiries to one tribe regardless of what visitors type, check that your tribes have their addresses placed on the map.
See it working
A franchise website running on Gorilla Dash
Hungry Gorilla is a demonstration website for a made-up cafe franchise. Its pages, location pages, menus, online ordering and enquiry forms are all managed in Gorilla Dash, so you can see how the pieces fit together before you build your own.
