Shoppers now ask AI assistants to find products for them. A person might type, "Find me a waterproof hiking jacket under $150 that ships in two days." The assistant then searches, compares, and sometimes even completes the purchase.
These agents do not care about your banner design or your homepage layout. They read your product data. If that data is thin, messy, or out of date, your products will be skipped, no matter how good your store looks.
For years, BigCommerce development work has focused on what shoppers see on the screen. The next stage focuses on what machines can read. This checklist helps you test your catalog against that standard, step by step.
What AI Shopping Agents Look For
An AI agent builds an answer from facts. It needs to know what a product is, what it costs, whether it is in stock, and how it differs from similar products. When a fact is missing, the agent either guesses or moves to a competitor with clearer data.
In practice, agents reward catalogs that are:
- Complete , with no blank attributes
- Consistent , with the same naming rules across products
- Structured , with facts stored in fields instead of buried in paragraphs
- Current , with accurate price and stock at all times
Keep these four words in mind. Every item below connects to at least one of them.
The 10-Point Catalog Readiness Checklist
The ten checks below cover the product data that AI agents depend on most. Each one is easy to test and can be fixed in a few clear steps. Go through them in order, and start with the checks that affect the largest number of products.
1. Unique identifiers on every product
Agents use identifiers to match your product with the same item across the web. BigCommerce lets you store SKU, UPC or EAN, MPN, and GTIN values. Fill in as many as apply, and make sure each one is correct.
A missing or duplicate identifier makes your product harder to verify. Start by exporting your catalog and sorting by blank identifier fields.
2. Titles that carry the key facts
A title like "Classic Jacket" tells an agent very little. A title like "Men's Waterproof Hiking Jacket, Navy, Size M" tells it a lot.
Write titles in a fixed pattern: brand, product type, key feature, color, and size. Keep them readable for humans, but make sure the main attributes appear near the front.
3. Attributes stored as fields, not paragraphs
Many stores hide important facts inside long descriptions. Material, weight, dimensions, compatibility, and warranty are all examples. A person can read them there, but a machine cannot rely on them.
Move these facts into custom fields or structured attributes. Use the same field names across the catalog, so "Material" never appears as "Fabric" on some products and "Made of" on others.
4. Variants and options set up correctly
Variants and modifiers serve different purposes. Variants are for options that change the SKU, price, or stock, such as size and color. Modifiers are for options that do not, such as engraving text.
When these are mixed up, agents may show the wrong price or stock status for a size or color. Review your top-selling products first and confirm that every sellable combination has its own accurate record.
5. Price and stock that match reality
An agent that recommends a product and finds it sold out will stop trusting your store. Price and inventory need to stay current, especially if they come from an ERP or warehouse system.
Check how often your sync runs and what happens when it fails. A nightly update may be too slow for fast-moving products. Add alerts so that a failed sync is noticed within hours, not days.
6. Clear shipping and returns information
Shoppers often ask agents about delivery time, shipping cost, and return rules. If this information is not available in a clear form, the agent cannot respond, and it may leave your product out of the result.
Publish shipping zones, delivery estimates, and return terms on dedicated pages with plain language. Where possible, keep them consistent with what the checkout shows.
7. Structured data on product pages
Structured data, also called schema markup, is a standard way to label facts on a page. It tells machines which text is the price, which is the rating, and which is the availability.
Built-in schema support is one of the most useful BigCommerce features for machine readability. Even so, a default setup may not cover everything. Test your product pages with a structured data testing tool, and confirm that price, currency, availability, brand, and review data appear correctly. If your theme has been customized, check that the markup was not removed.
8. Useful images with descriptive alt text
Agents that can read images use them to confirm what a product looks like. Even agents that cannot still rely on alt text and file names.
Write alt text that describes the product in simple words, for example "navy waterproof hiking jacket with hood, front view." Avoid stuffing keywords or repeating the same phrase on every image.
9. A consistent category structure
Categories help both shoppers and machines understand how products report. Overlapping categories, vague names, and products placed in many unrelated groups all weaken that signal.
Review your category tree. Each product should have one clear primary home. Use names that match how customers actually describe the products.
10. Reviews and trust signals
Agents often use ratings and review counts to rank similar products. Reviews also add natural language that describes real use.
Make sure review data is attached to the right product and appears in your structured markup. Remove spam or duplicate reviews, since low-quality data can hurt trust.
How to Score Your Catalog
You do not need to review every product at once. A sample gives you a good first picture.
- Pick 100 products. Include best sellers, new arrivals, and a few older items.
- Score each product against the ten checks above, using one point per check.
- Add the scores and divide by the number of products.
Use this table to read the result.
| Average score | Meaning | Next step |
|---|---|---|
| 9–10 | Strong foundation | Monitor and maintain |
| 7–8 | Good, with gaps | Fix the most common missing items |
| 4–6 | Uneven quality | Start a structured cleanup project |
| 0–3 | High risk | Rebuild data rules before adding new tools |
Look at patterns, not just totals. If most failures come from missing attributes, the fix is a data-entry rule. If they come from outdated stock, the fix is an integration change.
A 30-Day Improvement Plan
Small, steady steps work better than a large cleanup that never finishes. Here is a simple plan.
Week 1: Audit. Export the catalog, run your sample score, and list the top five recurring problems.
Week 2: Rules. Write short standards for titles, attributes, images, and categories. Share them with everyone who edits products, so new data stays clean.
Week 3: Fix the best sellers. Update the products that bring in the most revenue first. Improvements here show the fastest return.
Week 4: Automate and test. Use bulk import tools or scripts for repeated fixes. Re-run your sample score and compare it with Week 1.
After the first month, repeat the audit every quarter. Catalog quality slips over time as new products are added, and regular checks keep it from drifting.
Where to Start This Week
You do not need a large project to begin. Export your catalog, pick fifty products, and score them against the ten checks. That single exercise will show you where your data is strong and where an AI agent would struggle.
Fix identifiers, titles, and price accuracy first, since they affect the most queries. Then work through attributes and structured data. If your team needs outside help, Emizentech can run a catalog audit with you and turn the findings into a ranked list of fixes your developers and merchandisers can act on.
Agents will become a bigger part of how people shop, and the stores that prepare their data will now be easier for them to find, trust, and recommend.
