Writing Knowledge Assets for Chatbots

A quick guide to writing Knowledge Assets — Posts, FAQs, and Guidelines — so any chatbot (Brandkit's own, or a frontier model like ChatGPT or Claude) can find and use them accurately.

A few terms used below:

  • Knowledge Asset — a text-based Asset (e.g. Post, FAQ, or Guideline, etc).
  • Knowledge Library — the sum of all your Knowledge Assets.
  • Digital Asset — a file-based Asset (e.g. photos, logos, videos, PDFs, etc).
  • Asset Library — the sum of all your Digital Assets.
  • Brandkit — the overall system, bringing the Asset Library and Knowledge Library together under one domain and search engine.

Why format matters

A chatbot works by retrieving the most relevant Knowledge Asset (or the most relevant section of one) when someone asks a question, then using that content to answer.

If a Knowledge Asset is one long wall of text, the chatbot either pulls in too much (irrelevant info mixed in) or too little (cuts off mid-thought). Clear structure inside each Knowledge Asset fixes both.

Use markdown, not plain text

The Markdown format gives chatbots visible structure to work with — headers, lists, and emphasis all help them find and use the right piece of content. This is why we write in markdown: it's readable by Brandkit's internal chatbot and by external frontier model chatbots alike.

  • # and ## for headers — mark where one topic ends and another begins
  • - or 1. for lists — keeps facts scannable instead of buried in a paragraph
  • Bold for key terms, names, or numbers worth flagging

Brandkit's internal Writing tool (for Posts, FAQs and Guidleines) supports markdown natively and generates .md files automatically as you write — so following these formatting practices happens as part of normal editing, not as an extra export step.

One topic per Knowledge Asset, one sub-topic per section

Each Knowledge Asset should cover a single topic — one FAQ, one Guideline, one story. Within it, each section should stand alone: if someone retrieves just that section, it should make sense without needing the paragraph before or after it.

Good:

## Visitor Numbers
Around 110,000 people call Launceston home. The city sees
[X] visitors annually, with peak season in [months].

Not ideal:

## About Launceston
Launceston has 110,000 residents and a strong arts scene,
plus visitor numbers have grown 12% and there's a new
festival launching, oh and also here's our tone of voice...

(Too many unrelated facts crammed together — hard to retrieve cleanly.)

Lead each section with the plain-language summary

Put the core fact or answer in the first sentence of a section. Retrieval matches on relevance, so the opening line matters more than the closing one.

Keep static content separate from live content

Facts that don't change often (brand story, tone of voice, key history) belong in your Knowledge Library as Knowledge Assets. Photos, videos, and other media files belong in your Asset Library instead — chatbots reference them separately from the text-based Knowledge Library.

Facts that change often (event calendars, partner lists, visitor stats updated monthly) are better connected live — so a chatbot always has the current number instead of a stale Knowledge Asset from three months ago. Flag these to your Brandkit contact if you're not sure whether something should be a Knowledge Asset or a live connection.

A simple template

# [Knowledge Asset Title]
Source: [where this content comes from - you can use the Credit feature in Brandkit for this]
Last updated: [date] (This is taken care of automaticallu in Brandkit

## [Section Name]
[One-sentence summary of the key fact or answer.]
[Supporting detail, kept brief and self-contained.]

## [Next Section Name]
...

Quick checklist before publishing a Knowledge Asset

  • [ ] The Knowledge Asset covers one topic (one FAQ, one Guideline, one story)
  • [ ] Headers mark clear section boundaries within it
  • [ ] Each section makes sense on its own
  • [ ] No single section mixes more than one main topic
  • [ ] Key facts appear early in each section, not buried at the end
  • [ ] Content that changes often is flagged for a live connection instead of a static Knowledge Asset
  • [ ] Always add a summary - and you can auto-generate one with AI
  • [ ] Check any auto generated summary and copy to the Description field

Writing Knowledge Assets for Chatbots

Learn how to create Knowledge Assets (FAQs, guidelines, stories) for chatbots with markdown, one topic per asset, and self-contained sections. Separate static content from live data, use a simple template, and follow a checklist for accurate retrieval.

Asset type post
ID #886426
Word count 689 words

Licence

Licence: Worldwide Paid and Unpaid Available to anyone for royalty free use in paid and unpaid media worldwide, provided Brandkit benefits from such use, and Brandkit is credited (optional).
Expiry: No expiry date
Release date:
Added at:
Updated at:

Tags

Loading