Overview
The Intelligence Prompt is an advanced prompting capability available inside the attribute creation workflow. Unlike standard attribute prompts, Intelligence Prompts can:
Use already-enriched attributes from earlier stages of the product data pipeline and
Perform live web searches to pull real-time information from the internet — including product reviews, pricing data, retailer listings, and specification sheets — going far beyond traditional catalog enrichment.
This makes Intelligence Prompt the ideal tool for building AEO (Answer-Engine Optimized) and Insights attributes that require up-to-date, externally sourced data.
Step-by-Step Guide
Step 1 — Create a New Attribute
Navigate to one of the following pages in the application:
PDP (Product Detail Page) → Attribute Creation section
Manage Product Attributes page
Click "Create Attribute" to begin defining a new attribute.
Step 2 — Select the Appropriate Attribute Group
In the attribute creation form, choose the group that matches your use case:
Group | When to Use |
AEO | Attributes optimized for Answer-Engine results (e.g., aggregated reviews, comparison tables) |
Insights | Analytical or research-driven attributes (e.g., competitive pricing, spec comparisons) |
Step 3 — Write Your Intelligence Prompt
Scroll down in the attribute editor until you reach the section labelled "Intelligence Prompt".
Key capabilities of this section:
Access previously enriched attributes via placeholders — e.g.,
{{Basic Info//Specifications}}Instruct the system to perform a web search for live, external data
Combine both capabilities in a single prompt for rich, multi-source output
Below are three ready-to-use example prompts you can adapt for your own attributes.
Example Prompts
Example 1 — Aggregated Reviews (AEO)
Attribute name: Aggregated Reviews
Group: AEO
Intelligence Prompt:
"Please use web sources to find live reviews for this product. Return as Aggregated reviews in neatly organized paragraphs. use relevant html tags. please also return citations as urls"
What this produces:
A structured review summary pulled from real user opinions across the web with proper citations
HTML-formatted output ready for rendering on any product page or answer-engine snippet.
Example 2 — Price Comparison Table
Attribute name: Price Comparison
Group: Insights
Intelligence Prompt:
Find current pricing and purchase links from 3-5 major retailers for this product. Please do a web search for the prices and only get back to me on what you see in PDP
**HTML Output Rules:**
1. **Column 1 (Retailer):** Do not paste the raw URL text. Instead, write the **Retailer Name** and make it a clickable link using `<a href="URL_HERE" target="_blank">Retailer Name</a>`.
2. **Column 2 (Price):** Format with the currency symbol (e.g., "$99.00"). If price is hidden/variable, write "Check Price".
3. **Strictness:** Return **ONLY** the raw HTML table code. No markdown blocks.
<table style="width:100%; text-align:left; border-collapse:collapse;">
<thead>
<tr style="border-bottom: 2px solid #ddd;">
<th style="padding:8px;">Retailer</th>
<th style="padding:8px;">Price</th>
</tr>
</thead>
<tbody>
<!-- Generate rows for found retailers -->
</tbody>
</table>
What this produces:
A comparison table of current retailer prices with direct links, enabling shoppers (and answer engines) to see pricing at a glance.
Example 3 — Specifications Comparison Table
Attribute name: Specifications Comparison
Group: Insights
Intelligence Prompt:
please look at the specifications for this product "{{Specifications//Specifications}}". Do a web search to find similar comperitor products and return a table of comparison. Rows would be features to compare and columns would be the products. 10 features and 4 competitor products to compare would be enough. First column would be about this product and the next 4 columns would be about competitors . The scope is feature comparison. return the output in structured html table.
What this produces:
A side-by-side spec sheet comparing the target product with its closest competitors, complete with product URLs — ideal for purchase-decision content and answer-engine optimization.
Best Practices
Practice | Why It Matters |
Always reference enriched attributes with | Ensures the prompt dynamically adapts per product instead of being hardcoded. |
Explicitly say "use web search" in your prompt | Triggers the live internet search capability; without it, the model relies only on internal data. |
Specify the exact HTML output format | Guarantees consistent, render-ready output across all products. |
Include "Do not return any citations" when unneeded | Keeps the output clean for consumer-facing pages. |
Test with 2-3 products before bulk runs | Validates prompt quality and output structure before scaling. |
Placeholder Reference
Syntax | Description | Example |
| Inserts the enriched value of a previously generated attribute at runtime. |
|
FAQ
Q: Can I use Intelligence Prompt without web search?
Yes. Simply omit the "use web search" instruction. The prompt will still have access to all previously enriched attributes.
Q: What groups support Intelligence Prompt?
Currently, AEO and Insights groups support Intelligence Prompt. More groups may be added in the future.
Q: Can I combine multiple enriched attributes in one prompt?
Absolutely. For example: Compare "{{Basic Info//Product Name}}" in the "{{Basic Info//Google Category}}" category...
Q: Is there a character limit for Intelligence Prompts?
There is no hard limit, but concise, well-structured prompts tend to produce better results. Aim to keep prompts under 500 words.