AI product recommendations: what works, what doesn't, what's coming

We A/B tested every major recommendation engine on Shopify. Here are the ones worth the spend.

We A/B tested the major AI recommendation engines on Shopify across 6 stores and 4 categories. Here’s what actually moved revenue.

What works

The category-affinity engines beat the "customers also bought" engines on every store we tested. AI doesn’t need to be sophisticated — it needs to use the right inputs. PDP and cart recommendations driven by category affinity outperformed collaborative filtering by 18–32% in CR uplift.

What doesn’t

Homepage AI carousels. They look great in demos and underperform editorial-curated sections on every test. Cart-page AI bundles also underperformed our hand-curated bundles in 5 of 6 tests.

AI is best in places humans don’t scale: PDPs, cart recommendations, search relevance. Not where editorial taste matters.

Where it’s going

Generative AI for PDP description rewrites tested well — but only when paired with human editing. Pure-generated descriptions consistently lose. The 2026 unlock isn’t bigger models; it’s tighter feedback loops between AI output and conversion data.