How the AI Accesses Your Content
Choosing the right setup for public, private, live and user-specific content
Prezentare generală
Your chatbot answers using the content you give it. How you give it that content depends on two things: who is allowed to see it, and how often it changes. Public pages are simple, because the AI reads them and answers from them. It becomes more interesting when content sits behind a login, changes by the minute, or differs from one user to the next.
There are six ways to do this, and they can be combined. This page explains each one and helps you pick.
Which One Do I Need?
| What you want the AI to know | Use this |
|---|---|
| Public pages, documents and FAQs | Bază de cunoștințe |
| Pages behind a login, the same for everyone who logs in | Authenticated crawl |
| Information that changes constantly, such as stock or prices | Flux de date în timp real |
| Different content for different groups or plans | A separate chatbot per group |
| Details unique to each logged-in user | Contextul utilizatorului |
| Large or live per-user data from your own system | Instrumente personalizate |
1. Public Content
Everything you add to the knowledge base becomes available to the AI. You can add text, PDFs, documents, spreadsheets, videos and individual web pages from your dashboard, or let the crawler read your website automatically.
This is the right home for anything every visitor is allowed to see: product information, FAQs, policies, guides and general company information.
Read more about the knowledge base
2. Content Behind a Login
If your pages are only visible after signing in, the crawler can sign in too. Provide the address of your login page along with a set of credentials, and the crawler will log in and read the protected pages just as a logged-in user would.
This works well when every logged-in user is allowed to see the same material, such as a shared customer portal, an internal handbook or a members area with common resources.
Read more about the website crawl
Good to know: A knowledge base is shared by everyone who uses that chatbot. Content you add there can inform an answer to any of its users, so only add protected content when all users of that chatbot are allowed to see it. If access varies, use one of the options below.
3. Information That Changes Constantly
Some information is out of date as soon as it is saved: stock levels, prices, availability, opening hours during a holiday, waiting times. Rather than storing it, you can point the chatbot at a live feed.
You provide a URL on your own systems that returns the current data as JSON, CSV or plain text. The AI reads it when it needs it, so every answer reflects what your feed says at that moment.
The standard Real-Time Data Feed handles feeds up to 200,000 characters, which is roughly 500 items, and is available on Standard, Pro and Enterprise. For large catalogues, Real-Time Data Feed Max handles up to 10,000,000 characters, roughly 25,000 items, and is available on Pro and Enterprise.
A feed is shared by everyone using that chatbot, in the same way as the knowledge base. For data that belongs to one specific user, use User Context or Custom Tools instead.
Read more about Real-Time Data Feed · Flux de date în timp real Max
4. Different Content for Different Groups
When your users fall into a few clear groups that each get different material, the cleanest approach is a separate chatbot for each group, each with its own knowledge base. Your website then loads the chatbot that matches the signed-in user.
Because the knowledge bases are entirely separate, one group's content can never appear in another group's answers. There is no filtering to configure and nothing to get wrong.
This suits subscription tiers, partner levels, staff versus customers, or different departments and regions.
One account covers several chatbots: 1 on Free, 2 on Starter, 3 on Standard, 20 on Pro, and a custom number on Enterprise. You can add and manage them on the Manage Websites page.
Read more about managing websites
5. Details Unique to Each User
Some information belongs to one person only: their subscription, their order, their entitlements, their account status. That does not belong in a shared knowledge base at all.
With User Context, your website passes that user's information to the chatbot after your own system has authenticated them. You decide exactly what is included, and the AI uses it when answering. Because your server builds the context for each user separately, one user can never receive another user's information.
You are not limited to short values such as a role or a plan name. You can pass real content this way too, including a short document, as long as it fits within the limit. The context is capped at 2,000 characters per user on the Standard plan and 10,000 on Pro, with custom limits available on Enterprise.
Pass only what is needed. A role, a membership type or an entitlement flag is often enough for the AI to answer correctly, and keeps personal data out of the conversation entirely.
6. Large or Live Per-User Data
When there is more per-user material than the context allows, or the information has to be looked up at the moment it is asked for, the AI can fetch it from your own system during the conversation.
With Custom Tools, you define an endpoint on your server and describe when the AI should call it. The AI calls that endpoint and passes whatever it needs, such as an order number or the user identifier it received in the user context. Your server decides what to return, so access control stays with you and nothing has to be added to the knowledge base in advance.
This is the right approach for order status, account balances, personal records, and libraries of documents where each user is entitled to a different set.
Combining Approaches
Most setups use more than one. A typical membership site keeps its public pages and shared member resources in the knowledge base, passes the signed-in member's plan and entitlements through User Context, and uses Custom Tools for anything that has to be looked up live.
The AI draws on all of them together when it answers.
Where Your Content Is Stored
Content you add to the knowledge base is stored in your account so the AI can search it. Information that arrives through a real-time feed, through User Context or through Custom Tools is used to answer the message and appears in that conversation in your chat logs, but is never added to the knowledge base.
Conversations are kept so you can review them, and you can delete them at any time. Your content and conversations are not used to train AI models.