Grounding test cases in real shopper language, instead of internal guesses, is what makes the difference. Each prompt comes with a Volume estimate and an Opportunity Score, so you know exactly which prompts are worth optimizing for first when you're building out your Agentic Product Listings. That's how Prompt Libraries surface your demand gaps and get your Product Catalog ready to win with AI shopping agents.

The Prompt Libraries index, with status, prompt count and creation date for every library

The main Prompt Library page is your index of every library you've created. Each row shows:

Use the pagination controls at the bottom to browse your libraries. Click + New library in the top right to start the creation wizard.

New libraries come together in a four-step wizard: Name & describe, Add sources, Generate and Done.

Step 1, library name and description with the four-step progress indicator

Library name: a short, identifiable label used across the index and your tracked prompts.

Description: a plain-language brief on what you're testing for. Describe your product, target buyer personas and the topics you want covered, the way you'd brief a researcher.

Be specific here. This description shapes every part of how the library gets built, from the sources it looks at to the prompts it pulls out, so the more clearly you describe your product and buyers, the more relevant your results will be.

Click Next to move on to Add sources.

Step 2, three ways to add sources plus the keyword volume option

Sources are the raw material your prompts get built from. You have three ways to add them, and you can combine them:

Leave this checked to rank every prompt by real search demand and tie each one to a specific product to fix. Choose which country's shopper demand to measure against, or leave it on Auto to follow wherever your store ships. Each run draws on your search-data allowance.

Discovery results, filterable by community type

When discovery finishes, you get a panel of communities with a type tag, member count and a short note on why each one is relevant to what you described in Step 1. Filter by type (All, Buying Advice, Enthusiast Community, Lifestyle) to find the best fit, check the ones you want and click Done to add them to your source list.

Once at least one source is added, Generate library activates.

Click Generate library. Generation runs in the background with a real-time progress indicator, and the time it takes scales with how many sources you added. Feel free to navigate away. We'll keep working, and the library status updates to "Ready" when it's done.

When it's done, you get a confirmation screen with a Ready status badge. Click View library to see your results, or Create another to start a new one.

The Library Preview gives you full visibility into your generated prompts, with intent classification and demand metrics for each one.

The Opportunity map, plotting every prompt by volume against opportunity score and colored by funnel step

The Opportunity map plots your prompts by search demand against how well your catalog is currently showing up for them, colored by funnel step: Discovery, Research or Purchase.

Prompts in the top right are your best bets: real demand, real buying intent, and room for your catalog to close the gap. Use the map to spot which clusters to prioritize before you dig into the table below. Expand How each prompt is ranked for more on what feeds the score.

The prompt table, sorted by Opportunity from high to low

Prompts are sorted by Opportunity from high to low by default, so the biggest wins are already at the top. Each column tells you something different: