LiyaSupport
Getting started

Building your first knowledge base

Seed Liya with your content so AI drafts are accurate, grounded, and specific to your product from day one.

5 min read Updated July 2026

The knowledge base is the primary source of truth Liya draws on when drafting replies. You don't need a complete library before going live — even 10–20 focused articles covering your most common customer questions will produce noticeably better drafts than starting empty.

What to include first

Before writing any content, pull your last 30 days of support tickets and identify the top 10 questions by volume. These are the articles to write first. Common categories for most teams:

CategoryExample articles
Billing & paymentsHow to upgrade, cancel, or pause a subscription; refund policy; payment failure troubleshooting
Account accessReset password, change email address, manage 2FA, delete account
Getting startedInitial setup steps, onboarding checklist, key features overview
Product featuresHow specific features work, limitations, common configuration questions
Orders & shippingTrack an order, report a missing delivery, return & exchange policy
Technical / errorsCommon error messages and their fixes, system requirements, integration troubleshooting
Start with your highest-volume intent. If 40% of your tickets are billing questions, write five billing articles before anything else. One great article resolves hundreds of future tickets.

Adding your first articles

1

Open the Knowledge Base

In your dashboard, click Knowledgein the left sidebar. If this is your first visit you'll see an empty state with a prompt to add your first article.

2

Click 'Add article' and choose your input method

Three options are available:

  • Write / paste — type or paste content directly into the editor (supports Markdown)
  • Upload a file — PDF, .docx, .txt, or .md (up to 10 MB)
  • Add a URL — crawl any publicly accessible help page or doc

For your first knowledge base, pasting directly tends to give the cleanest results because you can control exactly what text is indexed.

3

Write one article per question

Create a separate article for each distinct question — don't bundle multiple FAQs into one long document. Short, focused articles produce far better AI retrieval than long multi-topic documents.

A good article structure:

  • Title: the question phrased as a customer would ask it (e.g. "How do I cancel my subscription?")
  • Body: the answer in plain language, 100–400 words
  • Include variations: mention alternative phrasings customers use (e.g. "downgrade", "unsubscribe", "stop plan")
4

Add topic tags

Tag each article with one or more topics: billing, shipping,account, technical, etc. Tags help Liya retrieve the right article when a conversation's intent overlaps multiple topics.

5

Save and wait for indexing

After saving, Liya indexes the article within 30–60 seconds. A green status indicator appears when it's ready for use in AI replies. You can add the next article immediately without waiting.

Importing content you already have

Most teams already have support content somewhere — in a Notion doc, a Google Doc, a Zendesk help center, or an internal wiki. Rather than rewriting from scratch:

  • Notion: Connect the Notion integration under Settings → Integrations and select which pages to sync. Liya imports automatically and stays in sync as you edit.
  • Google Docs: Connect Google Docs via Settings → Integrations for real-time sync.
  • Zendesk / Intercom / HubSpot Knowledge Base: Use the help center sync option. Liya crawls your public articles daily.
  • PDF manuals or policy docs: Upload via the file upload option. Liya extracts the text and splits it into indexable sections automatically.
  • Any public URL: Paste the URL and Liya crawls it. Works great for public help centres, changelog pages, and product documentation sites.

Quality checklist before going live

Before your team starts using Liya on real tickets, confirm:

  • Your top 10 most-asked questions each have at least one article
  • All articles are tagged with relevant topics
  • Articles are written in the same tone/language your customers use
  • Pricing, policy, and refund information is accurate and up to date
  • Each article is 400 words or fewer (split longer ones into sub-articles)
Don't wait for your knowledge base to be "complete" before going live. Start with what you have, watch the Gap Analysisreport to see what's missing, and add articles weekly. Teams typically reach stable AI quality within 2–3 weeks.