Product Catalog & Recommendations

Product recommendations need a catalog. AXO builds one for you from whichever sources you have – then serves AI-ranked product cards through product_reco zones. Review everything it knows on the Catalog page.

Getting products in

  • Shopify sync (zero effort). Link your store from the Shopify App Store install flow; products sync automatically through the store connection – no upload needed.
  • On-page scraper (automatic). The tag detects products on the pages visitors actually browse – via JSON-LD structured data, Open Graph meta tags, or DOM heuristics – and submits them at idle priority, throttled per URL. This keeps the catalog fresh with live names, prices, images, and availability even without any integration.
  • CSV / JSON import. Push your feed to POST /api/catalog/import with your site API key. Only name is required; price, category, URL, SKU, image, brand, description, availability, and ratings are all supported.

When sources overlap, the freshest wins: Shopify data takes precedence, then live scraped data, then imports, then AI inferences from the site crawl.

The Catalog page

Three views: Products (every unique product merged across sources, with image, price, category, availability, rating, and SKU), Categories (auto-extracted category groupings with product and page counts – products can be reclassified if the AI filed one wrong), and Pages (your pages grouped by type, each with its extracted product list and a semantic summary).

Recommendation zones

A product_reco variant renders a product card rail inside any zone. Three authoring modes:

Curated

You hand-pick the exact products and their order. No engine ranking – what you pin is what renders.

Auto-rank

The engine picks; you set the strategy – session_behavior (based on what this visitor browsed), bestsellers, new_arrivals, or similar (to the viewed item) – and the scope (full catalog, or restricted to one category). If a strategy has no data for a visitor (e.g. a fresh session), it degrades gracefully to bestsellers.

Mixed

Pin a few products to stable slots; the engine auto-ranks the rest up to your max. Best of both when merchandising rules meet personalization.

Each card is capped at a configurable max (up to 20 products) with a fallback behavior when nothing qualifies. On sites with semantic ranking enabled, products are also embedded and tagged with vibe codes – semantic descriptors the engine matches against the session's dominant vibe.

Preview before you publish

Use Preview as to see the live ranking for a fresh visitor, a random sample from a segment, or a specific session. The preview shows each recommended product with the engine's reason for choosing it, the strategy actually used (including any degradation), and the decision latency – so you know exactly what a real visitor in that state would see.