An Introduction to Prompting • Chapter 3

Context and Reference

The difference between a good response and a great one often lies in what you give AI to work with. Rich context and clear references transform AI from a general knowledge tool into a specialist—one that speaks your language, understands your needs, and delivers the results you’re looking for.

Estimated time.

Part 1

Providing Examples.

Examples are like showing the AI a prototype—they give it a tangible pattern to follow, reducing ambiguity and clarifying your expectations.

 

AI excels at recognising patterns. By providing examples you guide it toward the style, structure, and tone you’re looking for. Without them, AI might guess at your intent—sometimes hitting the mark, sometimes missing entirely.

 

Example:

❎ Vague Prompt: Write a scary story.

✅ With Example: Write a short scary story about a man alone in a cabin who discovers fresh footprints in the snow leading to his door. Here’s an example of the style I’m looking for:

“The old floorboards always creaked, but this sound was different. Sarah froze, coffee cup halfway to her lips, as footsteps echoed from her darkened hallway. Impossible—she lived alone, and her door was locked. Yet there they were again, slow and deliberate, coming closer. The lamp beside her flickered once, twice, then plunged her into darkness.”

“Using this same suspenseful style—strong sensory details, a slow build of tension—write the story.

💡 Why It Works:

✔ The AI now understands the tone, pacing, and level of detail you expect.

✔ It removes guesswork, ensuring the response aligns with your creative vision.

✔ Instead of just giving instructions, you show the AI what you want—leading to far better results.

Part 2

Providing References

Think of AI as a collaborator—its potential is vast, but its perspective is shaped by the knowledge it draws from. When you provide references, you give it context, ensuring that responses are not just accurate, but relevant to your needs.

References are especially valuable for overcoming one of AI’s biggest limitations: its training data has a cut-off date. It doesn’t know about recent developments, niche topics, or specialised information—unless you provide that input.

 

Two Types of References

 

1. General References

General references guide AI toward established knowledge—things it already “knows” . AI has broad knowledge, but without guidance, its responses can be too generic or unfocused. By providing references, you direct it toward specific concepts, frameworks, or perspectives you want it to use.

Example:

❎ Vague Prompt:

“Explain the key principles of minimalism in design.”

✅ With Reference:

“Explain the key principles of minimalism in design, referencing Dieter Rams’ philosophy and Apple’s early products.”

💡 Why It Works:

Instead of pulling from a wide range of design theories, AI focuses on Rams’ approach and Apple’s application of minimalism, making the response more relevant and precise.

2. Further References

These references introduce new information—data, research, or text that AI wouldn’t otherwise have access to. Think of it as handing AI a new source to work from.

Example 1: Incorporating New Research

✅ Prompt:

“Based on the findings in this report:

[‘In 2025, renewable energy adoption in urban areas grew by 30%, driven by advancements in solar panel efficiency and supportive government policies...’]

Summarise the key drivers of this growth and their potential impact on urban infrastructure.”

💡 Why It Works:

This feeds AI real-world data, allowing it to generate a response based on up-to-date insights, rather than outdated training information.

Example 2: Refining Creative Work

✅ Prompt:

“Here’s a draft product description for our app:

[‘GreenConnect links sustainable businesses with eco-conscious consumers. With smart recommendations and a seamless interface, it’s never been easier to shop green.’]

How can we improve this description to emphasise the app’s innovative features and appeal to younger audiences?”

💡 Why It Works:

By supplying a draft, you give AI a baseline to improve upon, ensuring its refinements align with your tone, messaging, and audience.

Part 3

Using Delimiters

Imagine writing a long email without using paragraphs or punctuation. It would be chaotic, hard to follow, and likely misunderstood. Just as punctuation organises your thoughts in writing, delimiters structure your prompts, ensuring the AI interprets each part of your input correctly.

 

Delimiters are little signposts—like triple quotes, brackets, or tags—that show AI where one part of your prompt starts and stops. They guide the AI, helping it parse complex instructions, separate tasks, and maintain clarity in its response.

 

Why Delimiters Matter

Delimiters play a crucial role in multi-part or detailed prompts by:

 

✔ Enhancing Clarity: They separate different elements of a query, ensuring the AI can identify and process each part independently.

✔ Improving Organisation: Delimiters create a logical structure, akin to breaking text into paragraphs or chapters.

✔ Boosting Accuracy: By isolating key sections, delimiters reduce ambiguity, helping the AI deliver more precise and contextually relevant responses.

 

Types of Delimiters and Their Use Cases

 

1. Triple Quotation Marks

Borrowed from programming languages like Python, triple quotes are ideal for enclosing multi-line text or inputs with internal punctuation. They are especially useful when a task requires the AI to focus on a specific block of text.

 

✅ Example prompt, summarising text:

Summarise the following text, delimited by triple quotes, in three bullet points.

“””Artificial intelligence is rapidly advancing across industries, from healthcare to education. In healthcare, AI improves diagnosis and personalises treatment. In education, it enhances learning experiences through adaptive technologies. However, ethical concerns about bias and data privacy remain significant challenges. “””

 

💡 Why It Works

Triple quotes clearly indicate the text block to be summarised. This keeps the AI focused, reducing the risk of it misinterpreting or including instructions as part of the task.

2. XML Tags

Inspired by web development and data structuring, XML tags can clearly separate multiple inputs, making them ideal for tasks involving comparison or multi-part analysis.

 

✅ Example prompt, comparing articles:

“You will be provided with two articles (delimited with XML tags) on the same topic. First, summarise the arguments of each article. Then, indicate which one presents a stronger case and explain why.

 

<article> The first article argues that renewable energy adoption is essential for urban sustainability... </article>

<article> The second article highlights the challenges of transitioning to renewable energy in densely populated areas... </article>”

💡 Why It Works

The XML tags clearly separate the two articles, helping the AI treat them as distinct pieces of text for individual analysis before comparing them.

 

3. Brackets and Other Symbols

For simpler tasks, brackets, colons, or other symbols can act as delimiters to separate key parts of your input.

 

✅ Example prompt, language translation:

“Translate the following text, enclosed in brackets, to French.

[I would like a table for two near the window, please.]”

 

💡 Why It Works

The brackets clearly isolate the text to be translated, ensuring the AI focuses only on that segment.

 

Delimiters are the unsung heroes of effective prompting. By creating structure and clarity, they ensure the AI interprets your inputs accurately, even in complex or multi-part tasks.

Reflection

The right references don’t just improve accuracy—they shape the AI’s entire perspective, transforming it from a generalist into a specialist.

But this isn’t just about AI. It’s about collaboration, shared knowledge, and how we build on each other’s ideas.

Think about the creative partnerships that drive real innovation. Would Apple have redefined design without Dieter Rams? Would the Wright brothers have achieved flight without studying birds and engineering principles? Every breakthrough is built on the foundation of what came before—and AI is no different.

So the question isn’t just what AI can do for you, but what you can give it to work with. What knowledge will you bring to the table? How will you shape the conversation?

Because in the end, the best collaborations—human or machine—start with the right foundation.

Next • Chapter 4

Mastering Complex Conversations

Complex thinking demands clear dialogue. By breaking down intricate challenges and guiding AI through deeper reasoning, you can turn overwhelming problems into structured, solvable steps. Big challenges don’t need simpler answers—they need better questions and instructions. Master this, and you’ll unlock AI’s true power: transforming complexity into clarity.