AI similar items
The AI Similar Items logic leverages OpenAI Embeddings to compare product titles and descriptions, identifying items with similar textual content. This powerful feature allows for immediate recommendation of lookalike products, even for new SKUs with minimal or no behavioral history.
How it works
- Transforms product catalog text into vector embeddings using OpenAI Embeddings, then identifies the nearest semantic neighbors.
- Recommendations are anchored to the product currently being viewed or within a relevant product context.
- Operates independently, without needing to wait for co-view or co-purchase graphs to mature.
Best placements
Product Detail Pages (PDPs) (Primary)
Ideal for a 'More like this' section, facilitating product discovery and offering substitutions.
Things to know
- Recommendation quality is directly influenced by the clarity and richness of product titles and descriptions in your catalog.
- Due to text similarity, recommendations might include color or size variants. Filters can be applied to refine these suggestions.
- This logic complements collaborative filtering: AI addresses the 'cold start' problem, while collaborative logics provide social proof.