How to find the PDPs failing in AI search.

The slow way: four tools and a spreadsheet

The manual version burns a week. Pull a Screaming Frog crawl to find PDPs missing schema. Export the Ahrefs site audit and flag thin pages. Pull Google Search Console for falling impressions. Then manually check your top ten SKUs in ChatGPT and Perplexity. Finally, open a spreadsheet and try to correlate four data sources that were never designed to agree.

Halfway through, they start contradicting each other. The crawler says the schema is fine; ChatGPT ignores the product anyway. Ahrefs flags your best-seller as "thin." GSC shows stable traffic on a page with zero AI citations. None of these tools measures the thing you care about — whether an AI engine cites the page — so you end up with a shortlist built on gut feel and fix the PDPs that bug you most. Better than nothing; not a ranked, diagnosed list of the SKUs actually losing revenue to AI answers.

The eCommerce Insights way

1. Run a full catalog scan. Connect the store (Shopify admin app or Partner auth; URL-based scanning for other channels). Every PDP gets two scores: a citation score across ChatGPT, Perplexity, Google AI Overviews, Gemini, Claude, and Copilot, and an agent-readability score for whether a shopping agent can parse the page. A 2,500-SKU catalog takes roughly 40 minutes end to end.
2. Sort by biggest drop. In the PDP ledger, sort by week-over-week delta or filter for "citation count fell to zero." Filter further by collection, tag, or revenue band to focus on what pays the bills.
3. Read the gap diagnosis per SKU. Each low score names its gaps: description under 100 words, missing aggregateRating, blocked AI crawler, no FAQ block, canonical pointing at a variant URL, empty metafields. Each gap is labeled with expected score lift, so you can see a fix's impact before doing the work. Methodology in the PDP Score docs.
4. Pick the top three to fix this week. The fix queue ranks by revenue contribution times expected lift. The top three are almost always right; override for campaign timing or a pending launch. The full prioritization logic is its own job: prioritize which PDPs to fix first.
5. Ship the diffs and rerun. Approve the recommended diffs — Shopify push on the Growth plan (early access), CSV export on every plan — and rerun after seven days. Fixes typically show in AI answers within two to three weeks.

For the conceptual backdrop — why AI engines skip pages Google ranks — read the product AI visibility guide and see PDP optimization for how the fix workflow runs end to end.

What "good" looks like

A healthy catalog in the eCommerce Insights ledger, per audit data as of early 2026 (illustrative):

If your first scan finds 40% of SKUs below 60, that is a finding, not a failure — most catalogs have never been audited against AI-answer criteria, and the first month of fixes delivers the biggest lift per hour of work. Validate hero-SKU schema in Google's Rich Results Test as a free cross-check.

Related jobs