How to find PDPs failing in AI search.

How people do this without eCommerce Insights

The manual version of this job burns a week. You pull the Screaming Frog report to find PDPs missing schema. Then you export the Ahrefs site audit and flag pages with thin content. Then you go to Google Search Console and pull the queries with falling impressions. Then you open ChatGPT and manually check whether each of your top ten SKUs is being cited. Then you open a spreadsheet and try to correlate the four data sources.

Halfway through, you notice the sources disagree. Screaming Frog says the schema is fine; ChatGPT is ignoring the product anyway. Ahrefs flags a "thin" PDP that happens to be your best-seller. GSC says traffic is stable on a page that has no AI citations. You can't tell whether the problem is schema, content, review signal, or something else entirely. The tools were built for SEO before AI answer engines existed, so none of them directly measure AI citation.

Most teams end up with a shortlist of suspicious PDPs based on gut feel, and fix the ones that bug them most. That's better than nothing; it is not a ranked, diagnosed list of the SKUs actually losing revenue to AI answers.

How to do this in eCommerce Insights

1. Run a full catalog scan. Connect your Shopify store (admin app install or Shopify Partner auth). eCommerce Insights reads every PDP, parses schema and metafields, queries ChatGPT, Perplexity, Google AI Overviews, Gemini, Claude, and Copilot for category-typical buyer prompts, and scores every SKU 0-100 on five weighted factors. A 2,500-SKU catalog takes roughly 40 minutes end to end.
2. Sort by biggest score drop. In the eCommerce Insights dashboard, sort the PDP list by week-over-week delta. Filter for "dropped more than 10 points" or "citation count fell to zero." The top ten rows are the PDPs losing ground fastest. Optionally filter by SKU tag, collection, or revenue band to focus on what matters most.
3. Read the gap diagnosis per SKU. Each PDP has a ranked gap list: missing aggregateRating, description under 100 words, no FAQ block, canonical pointing to a variant URL, metafield values missing, image alt text absent. Each gap is labeled with expected score lift so you can see the impact of a fix before doing it.
4. Pick the top three PDPs to fix this week. eCommerce Insights ranks suggested fixes by revenue contribution times expected score lift. The top three are almost always the right answer. Override manually if you have campaign timing, seasonality, or a pending launch that changes priority.
5. Apply fixes and rerun. Apply the recommended diffs. If you are on the Shelf plan, eCommerce Insights can push approved diffs to Shopify admin directly via the admin API (early access). Rerun the scan after seven days. The gap disappears from the list when the fix shows up in AI answers — typically within two to three weeks.

What "good" looks like

A healthy Shopify catalog in eCommerce Insights looks like:

• At least 70% of SKUs scored above 60. Anything below 60 is "work to do."
• At most 5% of SKUs scored below 30. Those are the "missing" tier — rebuild or delete.
• Zero PDPs with week-over-week drops of 20+ points. A 20-point drop usually signals a theme change, a review-data disconnection, or a content rewrite gone wrong.
• Weekly lift on at least 3 PDPs. A healthy optimization rhythm ships three fixes per week, and at least that many should show movement on the tracker.

If your first full-catalog scan finds 40% of SKUs below 60, that is a finding, not a failure. Most Shopify catalogs have never been audited against AI-answer criteria, and the first month of fixes typically delivers the biggest score lift per hour of work. Pair with the Product AI visibility guide for the full picture.

Related jobs

Related tools

• Shopify SKU Visibility Grader — five-SKU preview of the full catalog scan.
• AEO Grader — store-wide readiness score for the quarterly view.