Shopify Product Feed Optimizer: Fix What AI Agents Actually Check

The two-era problem in feed optimization

Feed optimization tools were built in the era of Google Shopping XML feeds. Their job: make sure your g:id, g:price, g:availability, and g:condition are correct so your products appear in the Shopping tab.

That still matters. But since March 2026, there's a second era: AI shopping agents that browse your store directly via /products.json, crawl your PDPs for structured data, and use LLM-based reasoning to decide whether your catalog is worth recommending. They're not reading your Merchant Center feed. They're reading your store.

The optimization problem is now two-dimensional: traditional feed compliance for paid channels, plus AI-agent readiness for the new organic channel. Most tools only solve dimension one.

What traditional feed optimizer tools cover

Tools like DataFeedWatch, GoDataFeed, Feedonomics, and Shopify's own Google & YouTube app optimize the XML/CSV feed that gets submitted to Merchant Center. Their audit checklist typically covers:

The overlap is thin: only GTINs and product images are genuinely shared priorities. Everything that makes AI agents recommend you — descriptions, structured data, review signals, category taxonomy — is outside the scope of traditional feed optimization.

The 7 things an AI-era feed optimizer needs to check

• 1

  robots.txt access gates

  If User-agent: GPTBot or User-agent: PerplexityBot is blocked in robots.txt, no amount of feed optimization matters — the agent can't see your store. This is the most common and most silent failure mode. Check it first.

• 2

  /products.json accessibility and completeness

  Your /products.json endpoint is what most AI agents use for bulk catalog discovery. It needs to be publicly accessible (no password protection), paginated correctly (250 products per page max), and include variant data. How to test your feed URL →

• 3

  GTIN coverage across variants

  GTINs (barcodes) are the product identifier that lets AI agents do price comparison, exact variant matching, and cross-retailer consolidation. Missing GTINs on a variant means that variant is effectively invisible to AI-powered price comparison. Coverage needs to be per-variant, not per-product. GTIN guide for Shopify →

• 4

  Description richness and specificity

  ChatGPT Shopping and Perplexity Commerce use your product description as citation material when answering a user's query. Thin descriptions (under 100 words, or just a size chart) score poorly. Rich descriptions covering use case, materials, fit, and features score well. This is the highest-leverage signal most stores get wrong.

• 5

  AggregateRating JSON-LD on PDPs

  Perplexity Commerce treats review count and average rating as a direct recommendation confidence signal. ChatGPT Shopping uses it as a secondary quality indicator. You likely have reviews in your app — but if they're not in JSON-LD on the product page, agents can't read them.

• 6

  ProductGroup JSON-LD for variant products

  Shopify products with multiple variants (size, color, material) need ProductGroup structured data that links parent product to individual Product items with their own GTINs. Without this, AI agents see one product instead of a full variant matrix — and can't answer "do you have this in medium?"

• 7

  Shopify AI Shopping category taxonomy

  Shopify's Agentic Storefronts (enabled by default since March 2026) use its own AI Shopping category taxonomy. If your products are categorized using old Shopify types or custom tags instead of the standardized taxonomy, they don't surface in the Shopify Global Catalog used by AI agents browsing via the Storefront API.

The optimization sequence: what to fix first

If you run an audit and find issues across all 7 areas, prioritize in this order — each fix unlocks the next:

1. robots.txt — 30 minutes; immediate unlock for all AI agent crawling
2. /products.json access — 15 minutes; confirms the bulk catalog endpoint works
3. AggregateRating JSON-LD — 2–4 hours with a review app theme integration; high leverage for Perplexity Commerce
4. ProductGroup JSON-LD + GTIN — 1–2 days for stores with complex variant catalogs; required for exact-match variant lookup
5. Description rewrites — highest effort, highest long-term payoff; prioritize top-20 best-selling SKUs first
6. Shopify AI Shopping categories — Shopify admin bulk-edit or metafield update; affects Shopify Agentic Storefronts specifically
7. FAQ schema on PDPs — adds citation surface for Perplexity; good for informational queries tied to your product type

How CatalogScan runs this audit automatically

CatalogScan is an AI-era feed optimizer: it scans your Shopify store across all 18 AI-readiness signals and returns a 0–100 score with the top-5 highest-leverage fixes. It reads your store directly — /products.json, your PDPs, your robots.txt, and your structured data — the same way AI shopping agents do.

The free scan takes 2 minutes and requires no app install. It checks:

• robots.txt access gates for all major AI crawlers (GPTBot, PerplexityBot, OAI-SearchBot, Googlebot)
• /products.json availability, pagination, and variant field coverage
• GTIN coverage across product variants
• Structured data presence and validity (Product, ProductGroup, AggregateRating, FAQPage)
• Description richness metrics (word count, specificity signals)
• AI Shopping category assignment
• Review signal accessibility

Pro tier adds bulk auto-fix: metafield fills, description rewrites via Claude, and GTIN enrichment from GS1 lookups — the same 7 categories covered above, automated.