Reference Material in Prompts

Reference material in prompts means giving the AI a source to use while answering. This source may be a document, article, policy, transcript, dataset description, brand guide, example, or previous draft.

Reference material helps make the answer more grounded. Instead of relying only on general knowledge, the model can use the supplied source to produce a response that fits the given material.

What is Reference Material?

Reference material is any content that the AI should use as the foundation for its answer. It may contain facts, rules, examples, tone, structure, or background that the model must follow.

Core Idea: Reference material gives the AI a source to work from instead of making it rely only on general patterns.

Types of Reference Material

Documents
Reports, chapters, policies, manuals, research notes, or meeting transcripts.
Examples
Sample outputs, writing style samples, templates, or previous successful responses.
Data Descriptions
Column descriptions, business definitions, metrics, dashboards, or dataset notes.
Brand Material
Brand voice, tone guidelines, product descriptions, audience personas, or messaging rules.

Reference Material vs General Context

Element Meaning Example
General Context Background that explains the situation. The audience is beginner students learning prompt engineering.
Reference Material Specific source content the AI should use. Use the following course outline as the base for this lesson.
Output Instruction The task to perform using the reference. Convert this outline into a 600-word lesson with examples.

How to Use Reference Material Properly

Make it clear whether the AI should use only the reference material or whether it can also use general knowledge. If accuracy matters, instruct the model not to add unsupported claims beyond the source.

Reference-Based Prompt Flow

Add Source
→
Define Task
→
Set Source Rule
→
Define Output
→
Review Grounding

Practical Reference Prompt

Example Prompt

“Use the reference material below to write a beginner-friendly explanation of context engineering. Do not add concepts that are not supported by the reference unless clearly marked as an additional note. Return the answer with headings, short paragraphs, and key takeaways.”

This prompt works because it tells the AI how strictly to follow the source and how to format the final answer.

Common Mistakes With Reference Material

Mistake What Happens Better Practice
No source rule The model may add outside claims. Tell the model whether to use only the source or combine it with general knowledge.
Unclear reference boundary The model may confuse instruction with source text. Use labels such as Reference Material and Task.
Too many references The model may mix unrelated information. Use only references that directly support the task.
No output format The answer may become a plain summary. Define the final deliverable clearly.

Important: When using reference material, separate source content from instructions so the AI does not confuse them.

High-Risk Mistake: Do not use confidential or private source material unless it is safe and necessary for the task.

[Image/Diagram: A source-grounded prompting visual showing reference material feeding into task instruction and producing a grounded output.]

Reusable Template

Reference Material Prompt Template

“Reference Material: [source]. Task: [what to do]. Source Rule: [use only source or combine with general knowledge]. Output Format: [format].”

Key Takeaways

  • Reference material gives the AI a source to use while answering.
  • It can include documents, examples, data descriptions, and brand material.
  • Always clarify whether the model should use only the source or add outside knowledge.
  • Separate reference material from instructions using labels.
  • Reference-based prompting improves grounding and consistency.