Skip to content
Academy
Marketing Academy · Field Work●SEO
All projects·From: SEO for Agentic Commerce
CoreTeardown· 50 minutes

Four Listings, One Agent: Teardown of AI-Shopping-Ready Product Pages

Chewy

Objective: Given four synthetic-but-realistic product listing specimens as an AI shopping agent would ingest them, identify which defects would cause an agent to exclude, down-rank, or mis-transact on each one, and separate real defects from plausible-looking non-issues.

You're auditing Chewy's pet-supplies feed ahead of onboarding to a second agentic commerce channel, reviewing four representative listing specimens pulled straight from the product data an agent would actually receive.

For each specimen, identify the defects that would actually break agent discovery or checkout, cite the defects, and don't get distracted by details that look wrong but aren't.

Before you start

What you'll need

Free path (everything below is enough to finish)

Google Sheets
FreeLog each specimen's defects and severity

Simple structured tracking, no setup

Google Search Console(optional)
FreeCross-check indexed product URLs when validating a real (non-synthetic) feed at your own company

Free, confirms which product pages Google has actually crawled and can compare against a feed export

The process

Specimens to review

This listing has strong schema in most respects. What's actually wrong with it, and what happens to it in an agent comparison?

Sample output
Product page JSON-LD for 'Blue Buffalo Life Protection Adult Dog Food, 30 lb':
{
  "@type": "Product",
  "name": "Blue Buffalo Life Protection Adult Dog Food, 30 lb",
  "gtin13": "0840243101234",
  "brand": { "@type": "Brand", "name": "Blue Buffalo" },
  "aggregateRating": { "@type": "AggregateRating", "ratingValue": "4.6", "reviewCount": "3120" }
  // no Offer object present anywhere in the markup
}

Specimen: synthetic, realistic

The feed and the on-page schema both have a GTIN. Is that enough?

Sample output
Feed row: SKU 'CHW-88213', GTIN '0186547203391', title 'Kong Classic Dog Toy, Large'.
Page schema.org markup: gtin13 '0186547203319' (transposed digits from the feed's GTIN).

Specimen: synthetic, realistic

Everything technical checks out. Would this listing win a recommendation slot against an established competitor?

Sample output
Newly-added listing, 'Chewy Exclusive Salmon & Sweet Potato Recipe, 24 lb'. Schema: Product + Offer both complete, GTIN present, price and stock current. aggregateRating: reviewCount '2', ratingValue '5.0'.

Specimen: synthetic, realistic

The schema says InStock. Is that trustworthy?

Sample output
Offer schema on a bestselling item: availability 'https://schema.org/InStock', price '$42.99'. Merchant Center feed sync log shows the last successful sync was 6 days ago; the item sold out on the warehouse system 2 days ago.

Specimen: synthetic, realistic

Final deliverable

A defect log across all four specimens, each defect tagged critical/moderate/cosmetic with a one-line fix owner (you vs. developer).

See a reference example
Sample output
HelloFresh, agentic-commerce teardown log (excerpt)

Specimen 2, Family Box protein swap SKU
  CRITICAL, developer: feed GTIN and page-schema GTIN mismatch on last digit, blocks cross-merchant matching

Specimen 4, weekly special listing
  CRITICAL, developer: stock schema says InStock, warehouse system shows sold out 3 days prior

Specimen 1, new box variant
  MODERATE, you: 5 reviews on a 4.9 rating, needs review-outreach campaign before agent visibility push

Success criteria

You're done when you can:

  • Correctly identifies all 4 planted critical/moderate defects across the specimens
  • Does not flag any of the 5 distractors as real defects
  • Assigns a plausible owner (you vs. developer) to each real defect