Skip to main content
This tutorial teaches you how to customize prompts to control how the AI generates content—from tone and style to what information it includes and how it’s structured.

​
When to Customize Prompts

Customize prompts when:
  • ✅ AI content is good but doesn’t match your brand voice
  • ✅ You want different tones for different product categories
  • ✅ Generated content misses important product details
  • ✅ You need specific format or structure
Don’t customize if: Default prompts are working well. Only customize when there’s a clear need.

​
Tutorial: Customizing a Description Prompt

Let’s walk through customizing the description attribute prompt for a fashion brand.

​
Step 1: Test with Defaults

Before customizing, test the default prompt:
1

Enrich 5-10 Sample Products

Use default prompts
2

Review Quality

What’s good? What’s missing? What feels off?
3

Identify Issues

Example findings:
  • ✅ Content is accurate and well-written
  • ❌ Tone is too formal (we want casual, friendly)
  • ❌ Missing styling suggestions
  • ❌ Doesn’t emphasize sustainability (our key differentiator)

​
Step 2: Access Prompt Configuration

1

Go to Products

Open any product
2

Navigate to Attributes

Click the Attributes tab
3

Find Description Attribute

Locate the “Description” attribute
4

Click Configuration Icon

Click the gear/settings icon next to it
5

Prompt Modal Opens

You’ll see three sections: Prompt, Constraints, Configuration

​
Step 3: Review Current Prompt

Look at the current configuration: Data Sources (what the AI references):
  • Product Title
  • Existing Description
  • Primary Collection
Instructions (how it writes): 
Generate a comprehensive product description highlighting
features, benefits, and use cases. Write in a professional tone.
Constraints:
  • Min Words: 150
  • Max Words: 200

​
Step 4: Modify Data Sources

Add more product context:
1

Click in Data Sources Field

Use autocomplete to add:
2

Add Relevant Fields

  • {{material}} - So AI can highlight fabric
  • {{color}} - To describe color accurately
  • {{brand}} - To mention brand when relevant
3

Updated Data Sources

Now: Title, Description, Collection, Material, Color, Brand

​
Step 5: Rewrite Instructions

Change tone and focus: New Instructions: 
Using the product {{title}}, {{material}}, and {{color}}, write
a friendly, approachable product description in 150-180 words.

Focus on:
1. How the product feels and fits
2. Styling suggestions (what to pair it with, when to wear it)
3. Our brand's commitment to sustainability and ethical production

Tone: Casual and conversational, like talking to a friend

Avoid:
- Overly formal language
- Generic marketing speak
- Technical jargon
Be specific about what you want AND what you don’t want. Both are equally important.

​
Step 6: Adjust Constraints

1

Set Word Count

  • Min Words: 150
  • Max Words: 180 (reduced from 200 for conciseness)
2

Format

  • Keep as: Plain Text

​
Step 7: Test the New Prompt

1

Click 'Test Prompt'

In the prompt configuration modal
2

Select Test Products

Choose 3-5 products to test with
3

Generate

Cernel creates test content using your new prompt
4

Review Results

Check:
  • ✅ Is the tone casual and friendly?
  • ✅ Does it mention styling?
  • ✅ Does it highlight sustainability?
  • ✅ Is it 150-180 words?

​
Step 8: Refine if Needed

If test results aren’t perfect, iterate: Common refinements:
Add to instructions: “Use contractions (it’s, you’ll, we’re) and write as if speaking directly to the customer.”
Add to instructions: “Always mention quality and how it affects comfort or durability.”
Add to instructions: “Vary sentence structure. Avoid repeating the same phrases across products.”
Adjust constraints: Make the range narrower (e.g., 160-170 words)

​
Step 9: Save and Apply

1

Click 'Save'

Your custom prompt is saved
2

Optional: Save as Template

Click “Save as Template” to reuse for other attributes or collectionsName it: “Description - Fashion - Casual Tone”
3

Enrich Products

Now all future description enrichments use your custom prompt

​
Example Prompts for Different Scenarios

​
Luxury Brand (Formal, Elegant)

Data Sources: {{title}}, {{material}}, {{brand}}, {{collection}}

Instructions:
Write an elegant, sophisticated product description emphasizing
craftsmanship, heritage, and exclusivity. Use refined language
that reflects the luxury positioning of {{brand}}.

Highlight the premium {{material}} and superior construction.
Include subtle references to timeless style and investment value.

Tone: Sophisticated yet accessible. Avoid pretentious language.

Length: 180-220 words

​
Technical Products (Specs-Focused)

Data Sources: {{title}}, {{features}}, {{specifications}}, {{compatibility}}

Instructions:
Write a clear, technical product description focusing on
specifications and capabilities. Lead with key features,
then explain use cases and compatibility.

Use bullet points for specifications. Be precise and factual.

Tone: Professional and informative. No marketing fluff.

Length: 150-200 words + bullet list

​
Kids Products (Simple, Fun)

Data Sources: {{title}}, {{color}}, {{age_group}}

Instructions:
Write a fun, simple product description that parents will
appreciate and kids will love. Emphasize safety, durability,
and how much fun kids will have.

Use simple, cheerful language. Mention age appropriateness.

Tone: Upbeat and reassuring. Parent-friendly.

Length: 100-150 words

​
Collection-Specific Prompts

Create different prompts for different product categories:
1

Set Base Prompt

Configure a default prompt at organization level
2

Override for Collections

  • Luxury Collection: Elegant, sophisticated tone
  • Budget Collection: Value-focused, practical tone
  • Kids Collection: Fun, safety-focused
3

Cernel Auto-Selects

When enriching, the most specific prompt is used automatically
Learn more about hierarchical prompts →

​
Best Practices

Always test custom prompts on 5-10 products before bulk enrichment
Include example output: “Example: ‘This cozy fleece hoodie wraps you in warmth…’”
Don’t say “professional”—say “professional but approachable, like a knowledgeable salesperson”
Prompt engineering is iterative. Test, refine, test again.
Build a library of proven prompts for different categories and attributes

​
Troubleshooting

Cause: Instructions may be contradictory or too vagueSolution: Simplify. Focus on 2-3 key points. Be direct.
Cause: Product data quality varies (some have good info, others don’t)Solution: Improve source product data OR adjust prompt to handle missing data gracefully
Cause: Test products had complete data; production products are missing key fieldsSolution: Make data sources optional or enrich attributes incrementally (e.g., enrich color first, then descriptions)

​
What’s Next?

Custom Templates

Save and reuse prompt configurations

Attribute Prompts Reference

Deep dive into prompt components

Bulk Operations

Apply custom prompts at scale

Managing Taxonomies

Organize products for better context
Need help? Email [email protected] with your prompt and we’ll help refine it!