How Your Bot's Knowledge (Indexes) Work

Your AI bot answers from your knowledge, and that knowledge is organized into indexes. This guide explains what an index is, how to pick the right one for a bot, and how to add more.

What an index is

An index is a named collection of your content that the bot can search — for example your website pages, your help articles, your product catalog, or an uploaded document set. You always have at least one index, and to you that index simply is "your knowledge."

Behind the scenes all of your indexes live inside a single secure container on your account. You never interact with that container directly — you only ever work with your named indexes. Adding content always adds it to one of your indexes, never to the raw container.

All Knowledge vs. a specific index

When you set up a bot you choose what it can search:

• All Knowledge (recommended default) — the bot searches everything across all your indexes. This is the safest choice: the bot is grounded in your full content and won't answer "I don't have information on that" for things you've actually added.
• A specific index — point the bot at one index (e.g. just your product catalog, or just your help articles) when you want a focused assistant that only speaks to that topic.
• A group of indexes — select several indexes so a bot draws on a curated subset (e.g. help articles + product catalog for a support-and-sales bot).

One exception: troubleshooting indexes (manuals and procedures, part of the Troubleshooting Knowledge add-on) are left out of "All Knowledge". Procedures stay out of everyday product answers on purpose. A bot reaches them through a workflow AI step that lists the troubleshooting index by name. See the Troubleshooting Knowledge guide.

A bot pointed at an empty index will answer ungrounded ("limited information") — so if a bot isn't answering from your content, check that its index actually contains documents.

Adding knowledge

You add to an index by:

• Crawling your website — the bot ingests your public pages into a website index.
• Syncing your help center / KB articles.
• Connecting a store (Shopify, BigCommerce) — products become a product index.
• Uploading files (PDF, DOCX, TXT).

New content is searchable within a few minutes of finishing import. If you re-import, unchanged content is skipped automatically, so re-imports are cheap and safe.

Website pages, uploads, and custom content

On the Knowledge tab, the source list shows the real content types your bot can search: Website Pages, Uploaded Documents, Custom Content, KB Articles, YouTube, and your connected stores. Turning on Website Pages also includes Uploaded Documents and Custom Content, and those two rows show "Included with Website Pages" until you turn Website Pages off. Your named indexes are listed below the sources, with a tag on any troubleshooting index or manual catalog.

Manuals and procedures

If your customers ask how to fix or set up your products, load the manuals into their own troubleshooting index instead of your general one. That is the Troubleshooting Knowledge add-on. It keeps each procedure together and keeps manuals out of product answers. See the Troubleshooting Knowledge guide for what to load and how scraping scope works.

Good practice

• Start a new bot on All Knowledge so it's grounded from day one, then narrow to specific indexes only if you want a focused assistant.
• Keep at least one populated index — an empty index is the #1 cause of "the bot doesn't know things it should."
• Re-import a source any time its content changes; the bot picks up the refreshed content automatically.