> ## Documentation Index
> Fetch the complete documentation index at: https://docs.armin.cx/llms.txt
> Use this file to discover all available pages before exploring further.

# Classic AI sources

> Add the knowledge your Classic AI setup uses to answer customers.

Classic AI sources are the factual knowledge base for the legacy AI setup. Add sources before you enable automatic replies in Classic AI.

<Warning>
  Move existing Classic AI automation to AI Agents over time, not only new work. Follow the [migration guide](/ai-agents/migrate-from-classic-ai) to transfer knowledge safely and retire overlapping Classic AI behavior once the replacement is proven. Start with [Getting started](/ai-agents/getting-started), and use the [Classic AI agent list](https://armin.cx/app/_/ai/agents) only during migration or for necessary source maintenance.
</Warning>

## Choose a source

| Source   | Best for                                                                                                 |
| -------- | -------------------------------------------------------------------------------------------------------- |
| Website  | Public product pages, help centers, shipping policies, return policies, campaigns, and store information |
| Files    | Stable internal knowledge in PDF, DOCX, CSV, or TXT format                                               |
| Q\&A     | One repeated customer question with one clear, stable answer                                             |
| Guidance | Policies with multiple conditions, exceptions, or judgment calls                                         |

## Add website content

Website sources accept a root domain, sitemap, or specific page. With Smart Sync enabled, Chatarmin CX checks them weekly for updates.

<Steps>
  <Step title="Open Sources">
    Open the [Classic AI agent list](https://armin.cx/app/_/ai/agents), choose your agent, then go to **Sources** → **Website**.
  </Step>

  <Step title="Add the website">
    Add a root domain such as `https://example.com`, a sitemap such as `https://example.com/sitemap.xml`, or a specific help-center section.
  </Step>

  <Step title="Review indexed pages">
    Remove pages the AI should not use and add missing pages manually.
  </Step>

  <Step title="Test after the first sync">
    Ask questions that should use the website and confirm the answers.
  </Step>
</Steps>

<Frame>
  <img src="https://mintcdn.com/chatarmincom/dklE7IAeCxDC1E8J/images/Bildschirmfoto2025-12-04um22.10.23.png?fit=max&auto=format&n=dklE7IAeCxDC1E8J&q=85&s=249913daaeb43f7bfc85cca33b6c2107" alt="Classic AI website source with indexed pages" width="2902" height="806" data-path="images/Bildschirmfoto2025-12-04um22.10.23.png" />
</Frame>

## Add files

Use files for stable knowledge that is not available on a public page, such as compatibility tables, sizing guides, warranty details, or internal fulfillment rules.

<Steps>
  <Step title="Open Files">
    In your existing Classic AI agent, go to **Sources** → **Files**.
  </Step>

  <Step title="Upload a focused file">
    Upload a PDF, DOCX, CSV, or TXT file. Keep one topic per file.
  </Step>

  <Step title="Test the content">
    Ask questions the file should answer before relying on it for automatic replies.
  </Step>
</Steps>

<Note>
  Large or mixed-topic files can reduce answer quality. Split them by topic before uploading, or contact support for help preparing a large knowledge base.
</Note>

## Add Q\&A

Use Q\&A for repeated questions with one clear answer. For example:

* How long does shipping take to Germany?
* Can I return a worn item?
* What size is Medium in EU sizing?
* Do you ship internationally?
* Is this promotion code still valid?

<Steps>
  <Step title="Open Q&A">
    In your existing Classic AI agent, go to **Sources** → **Q\&A**.
  </Step>

  <Step title="Write the question">
    Phrase it as a customer would.
  </Step>

  <Step title="Write the answer">
    Keep it direct, complete, and ready to reuse.
  </Step>

  <Step title="Test variations">
    Ask the same question in a few different ways and confirm the AI finds the answer.
  </Step>
</Steps>

## Add guidance

Use guidance when a customer question depends on a broader policy rather than one fixed answer. Good topics include return conditions, sizing, payment methods, loyalty programs, and delivery timelines.

<Steps>
  <Step title="Open Guidance">
    In your existing Classic AI agent, go to **Sources** → **Guidance**.
  </Step>

  <Step title="Choose one policy topic">
    Create a focused entry such as `Return policy` or `Size guide`.
  </Step>

  <Step title="Explain the decision">
    Include the policy, relevant conditions, exceptions, and when Classic AI should not answer.
  </Step>

  <Step title="Test real variations">
    Test customer messages that should use the guidance and edge cases that should not.
  </Step>
</Steps>

## Best practices

* Prefer a sitemap when your site has many pages.
* Remove expired campaign pages and old file versions.
* Keep files short and topic-specific. Split large, mixed manuals by topic.
* Keep one customer question per Q\&A entry and use customer language.
* Maintain policy facts in one source instead of duplicating them.
* Use guidance when an answer depends on several conditions.
* Test the AI on real tickets before turning on auto-reply.