How AI Recommendation Engines Decide What to Show You

"You might also like" isn't magic — it's a set of fairly well understood algorithms. Here's what's actually driving those suggestions, and how to get better ones.

The two basic approaches

Most retail recommendation engines lean on a mix of two techniques. Collaborative filtering looks at what people with similar browsing or purchase history bought, and suggests items from that pattern — "shoppers who bought this headset also bought this dongle." Content-based filtering instead looks at the attributes of the product you're viewing (brand, category, price band, specs) and finds similar items. Most modern engines blend both, often layering in machine-learning models that weight recent behaviour more heavily than old browsing history.

What signals actually feed the model

Recommendation systems typically draw on a combination of:

• Your recent browsing and search history on that site
• Items you've added to cart or wishlisted, even if not purchased
• What similar shoppers ultimately bought
• Current stock levels and margin — retailers often nudge in-stock, higher-margin items
• Broad demographic or location signals, such as being in New Zealand

That last point matters: recommendations aren't purely about relevance to you, they're also shaped by what the retailer wants to move.

Why this matters for price comparison

Because recommendation engines are partly commercially driven, the "recommended" or "top pick" badge on a retailer's own site isn't a neutral signal — it can reflect stock clearance or margin as much as genuine quality. That's one reason it's worth checking prices across multiple retailers rather than trusting a single site's suggested alternative, especially for higher-cost electronics like laptops or TVs.

Getting better recommendations

If you want more useful suggestions, staying logged in while browsing helps the engine build a clearer picture of your preferences. Being specific in search queries (model numbers rather than vague terms) also tends to produce tighter, more relevant results. Conversely, if you want to reset a recommendation feed that's gone stale or is stuck suggesting things you already bought, clearing site cookies or browsing privately for a while is usually the quickest fix.

The takeaway

AI recommendations are a genuinely useful discovery tool, but they're optimised for engagement and conversion, not necessarily for finding you the best deal. Treat them as a starting point for ideas, then verify pricing and specs independently before buying.