AI Prompting Best Practices

Overview

AI prompting determines how Halda generates responses for students. A well-written prompt ensures AI output is:

• Clear and accurate
• On-brand and student-friendly
• Consistent across submissions
• Aligned with institutional goals

This article outlines best practices for writing prompts that deliver reliable, high-quality responses without over-engineering or unpredictable behavior.

---

Start with the Outcome, Not the Tool

Before writing a prompt, ask:

• What should the student walk away understanding?
• What action should they feel confident taking next?

Prompts work best when they are outcome-driven, not feature-driven.

Avoid prompts like:

“Write a compelling response about the program.”

Instead, anchor prompts to student value:

“Help the student understand whether this program fits their goals and what their next step should be.”

---

Be Explicit About Tone and Voice

Never assume the AI knows how you want it to sound.

Recommended tone guidance

• Warm
• Supportive
• Conversational
• Clear and professional
• Student-first (not sales-heavy)

Example

“Write in a friendly, reassuring tone, as if speaking one-on-one with a prospective student.”

Avoid vague descriptors like engaging or exciting without context.

---

Tell the AI Who It Is Speaking To

Context dramatically improves relevance and accuracy.

Include details such as:

• Prospective vs. admitted student
• Undergraduate vs. graduate
• Career-switcher vs. early-career
• Working professional vs. full-time student

Example

“Assume the student is a working professional exploring an online graduate program.”

---

Use Structure to Control Output

AI performs best when given clear structure.

Helpful structural cues

• Paragraph limits
• Bullet point usage
• Section order
• Content hierarchy

Example

“Respond with a short introductory paragraph followed by 3–5 bullet points highlighting key program benefits.”

Avoid asking for long, unstructured responses.

---

Reference What the AI Already Knows

Halda’s AI is already aware of:

• Form questions
• Answer choices selected
• Conditional paths
• Program context

You do not need to restate everything—but you should tell the AI how to use that information.

Example

“Use the student’s selected program and stated motivation to tailor the response.”

---

Avoid Hard-Coding Facts When Possible

Whenever feasible:

• Describe concepts instead of listing exact figures
• Focus on support, structure, and outcomes

This reduces:

• Maintenance overhead
• Risk of outdated information
• Inconsistent responses across forms

If exact details are required, clearly instruct the AI to reference approved content only.

---

Set Guardrails for What Not to Do

Explicitly telling the AI what to avoid improves consistency and compliance.

Examples

• “Do not mention tuition amounts.”
• “Do not make admissions decisions.”
• “Do not guarantee outcomes.”
• “Do not reference internal processes.”

---

Keep Prompts Focused and Concise

Long prompts are not better prompts.

Best practices

• One primary objective per prompt
• Clear, direct instructions
• Minimal redundancy

If a prompt is doing too much, consider:

• Breaking it into structured sections
• Using traditional (non-AI) content where AI is unnecessary
  
  

---

Encourage Helpfulness, Not Hype

Students respond better to clarity than marketing language.

Prefer

• “Here’s what students typically experience…”
• “This program may be a good fit if you…”

Avoid

• Overly promotional claims
• Buzzwords without explanation
• Aggressive conversion language

---

Pair Prompts with Strong Defaults

Even great prompts work best when paired with:

• Clear top paragraphs
• Strong, action-oriented CTAs
• AI Search for long-tail or follow-up questions

AI prompting should enhance, not replace, intentional content design.

---

Test and Refine Iteratively

After publishing:

• Submit multiple test responses
• Vary inputs intentionally
• Review edge cases
• Adjust prompts based on real output

Small prompt changes often produce meaningful improvements.

---

Common Prompting Mistakes

• Writing prompts that are too generic
• Over-specifying every detail
• Forgetting to define tone or audience
• Asking AI to make decisions it shouldn’t
• Treating AI like a copywriter instead of a guide

---

Prompt Guidelines (Recommended)

Use the guidelines below to produce consistent, predictable, and on-brand AI responses.

Structure Your Prompt Clearly

Whenever possible, define:

• Style guide
• Response format
• Required sections
• Section-level instructions
• Approved links to reference

The more clearly structured the prompt, the more reliable the output.

---

Be Direct with the AI

Use command language.

Examples:

• “Do not include…”
• “Only respond to the selected answer option.”
• “End this section with the correct link.”

Clarity improves accuracy.

---

Use Links Strategically

When exact information matters:

• Include a direct URL paired with the relevant answer option
• This reduces the risk of hallucination and ensures accuracy

---

Styling and Formatting Support

The AI understands HTML and CSS.

• You may provide inline CSS to control:
  
  • Text color
  • Background color
  • Font weight

• Use this sparingly to reinforce brand consistency and hierarchy

---

Advanced Use Cases

Prompts can:

• Evaluate answers like a quiz
• Recommend personas or pathways
• Deliver conditional recommendations based on multiple inputs

For advanced or high-impact form builds, work with your Account Manager to design and validate these experiences.

---

Example Prompt (Recommended Pattern)

---

Final Thought

Strong AI prompting is less about creativity and more about clarity.

When prompts are clear, intentional, and student-centered, AI becomes a reliable extension of your enrollment strategy—not a wildcard.