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

# Reference Materials

> Ground your content in real data. If you didn't provide it, it doesn't get invented.

# Reference Materials

If you didn't provide it, it doesn't get invented.

Reference materials give DraftLift factual grounding. Upload documents, paste URLs, or add text directly — the AI uses them as primary sources during generation. No hallucinated statistics. No fabricated quotes. No "studies show" without an actual study.

## Supported formats

| Format            | How it works                                                  |
| ----------------- | ------------------------------------------------------------- |
| **PDF documents** | Upload PDFs — DraftLift extracts text content automatically   |
| **Web URLs**      | Paste a URL — DraftLift fetches and extracts the page content |
| **Plain text**    | Paste text snippets directly for quick reference              |

## Adding reference materials

1. Navigate to **Reference Materials** in the sidebar
2. Click **Add Reference Material**
3. Choose your format (document upload, URL, or text)
4. Add a title for easy identification
5. Save — the material is now available for use in generation

Search and filter your reference library by title or content as it grows.

## Using references in generation

Select reference materials during **Step 1** of the generation pipeline. You can choose multiple references to provide comprehensive context.

When references are included:

* The AI treats them as **primary sources** — it draws facts, data, and quotes from your materials
* It **does not invent** statistics, fabricate quotes, or hallucinate data points not present in your references
* When the vessel supports citations, **source URLs are linked naturally** within the text
* Reference content combines with your [memories](/features/memories) and [vessel](/features/templates) rules to produce output that's both factually grounded and voice-consistent

## Getting the most from references

* **Use specific sources.** A research report or detailed article gives the AI rich material. Generic references produce generic output.
* **Combine with memories.** References provide the facts. Memories provide your perspective. Together they produce content that's informed and authentic.
* **Keep references current.** Outdated data leads to outdated content.
* **Use multiple references.** For comprehensive content like blog posts, several complementary references give the AI a broader factual foundation.