Skip to main content
AI Shopping can only track what your prompts cover. The prompts you choose determine which products, competitors, and buying intents you can analyze. This guide walks you through how to build an effective prompt set for AI Shopping. It’s designed for brands that sell their own products; multi-brand retailers and marketplaces require a different approach and aren’t covered here.
If you’re new to Peec AI, we recommend starting with Setting up your prompts and Choosing the right prompts before continuing with this guide.
Two types of prompts are used throughout this guide:
  • A category prompt: Asks about a type of product, for example, “What are the best running shoes for beginners?” It lets you track how all products in that category perform.
  • A product prompt: Focuses on one specific product, for example, “Is the Nike Pegasus good for flat feet?” It gives you a more detailed view of how that individual product performs.

​
Step 1: Starting with categories

Every product in your catalog is tracked automatically. Peec AI matches your products to relevant category prompts, so a prompt like “best running shoes for beginners” tracks the performance of all the running shoes in your catalog, rather than focusing on a single product. Categories are the foundation of an effective AI Shopping prompt set for two reasons:
  • Every product is covered: Products are automatically matched to relevant prompts within their category, regardless of how those prompts are worded.
  • Your data stays organized: A clear category structure reflects how your product range is organized, making your AI Shopping data easier to analyze.
Use your categories as the structure for your prompt set, then build prompts around the different buying intents you want to track.
You don’t need a prompt for every product. Peec AI automatically matches each product to the category prompts it’s relevant to, so a prompt like “best moisturizers for dry skin” measures all the moisturizers in your catalog at once.This means a well-designed set of category prompts can give you product-level insights across your catalog without creating individual prompts for every product.

​
Step 2: Build each category’s core set

For each category, start with a focused set of prompts that covers the most important shopping decisions. Prioritize non-branded discovery prompts, then add a smaller selection of branded prompts. Non-branded discovery should make up most of your prompt set. These prompts show whether AI recommends your products when the shopper hasn’t mentioned a specific brand.
  • Start with a broad recommendation prompt, such as “What’s the best [product category] for [key use case]?”
  • Add variations for the most important use cases, needs, and buying criteria in that category. Refer to Step 3.
Branded prompts should make up a smaller portion of your set. They’re useful for understanding how AI responds when shoppers already know your brand, but because your brand is named directly, they’re more likely to include it and can increase your overall visibility. Aim to keep them to around 15–20% of your prompt set.
Peec automatically classifies these prompts as Branded, so you can filter for Non-branded at any time to focus on your discovery performance. There’s no need to create a separate project.
  • “What are the best alternatives to [your brand]?” Captures shoppers who already know you and are comparing options.
  • One or two comparisons against your top competitors, phrased as real buying decisions: “I’m training for my first marathon, should I choose [your product] or [competitor product]?” Comparisons often produce a side-by-side table, which is a good source of attribute data.
Where product prompts fit: Add product-specific prompts on top of your category prompts when you want deeper insights into individual products.
  • Small catalogs (up to about 100 products): Adding roughly three prompts per product can be manageable alongside your category prompts.
  • Hero products: Prioritize your most important products with prompts around comparisons, buying decisions, and other key purchase considerations.
  • Long-tail products: Add a smaller selection of prompts for products outside your main priorities. Group them under a single tag so you can track their performance without overwhelming your prompt set.
For large catalogs, avoid creating a prompt for every product. Each product is already matched to relevant category prompts, so individual prompts would mostly duplicate the coverage you already have. Instead, reserve product-specific prompts for your hero products where you want deeper insights.

​
Step 3: Think in dimensions

Once the core is in place, expand coverage across three dimensions: To size your prompt set, multiply the main dimensions:
Use this as your starting point for building a solid baseline. From there, expand your set by adding distinct buying situations, such as new buyer contexts, use cases, competitor comparisons, markets, or categories. If you have room to track more prompts, use that capacity to broaden your coverage rather than creating multiple variations of the same question. As long as each prompt represents a genuinely different buying situation, it can add useful signal. To make this framework work effectively, follow three key rules:
  • Prioritize distinct buying situations: A new intent, use case, or buyer context adds more value than simply rewording an existing prompt. For your most important prompts, however, tracking a few variations can be useful. Individual prompts can fluctuate from day to day, while several variations can give you a more stable view of performance.
  • Name something shoppable in every prompt: Include a specific product or product category rather than only describing a general need. For example, “Best Pilates socks for beginners” is more likely to trigger shopping results than “What do I need for my first Pilates class?”
  • Prompts with commercial intent: Shopping recommendations appear much more often for commercial prompts than informational ones. Keep some informational prompts where they’re useful, but focus most of your set on product discovery, comparisons, and purchase decisions.
Not every prompt will trigger a shopping recommendation, and that’s expected. Shopping results are more likely to appear for clear, product-focused queries, particularly for physical products rather than services.