How to build content recommendations in Bubble

The simplest way to build content recommendations in Bubble is to save what each user interacts with (views, likes, categories they prefer), store those as fields in the database, and then show content using a Repeating Group with a search that sorts items by relevance. Bubble doesn’t have built‑in “AI recommendation logic,” but you can recreate practical recommendation behavior by tracking user signals and then filtering/sorting based on them.

Core Idea

You create a User Profile of preferences, update it whenever the user interacts with content, and then use a Repeating Group search that sorts items by how many preference fields match.

• User stores fields like: viewed_categories (list of texts), liked_content (list of Content items).
• Content stores fields like: category (text), tags (list), popularity\_score (number).

Step-by-Step Structure

Data setup: Add fields on User that capture behavior. Each time a user views or likes a piece of content, run a workflow that updates their lists, for example “User’s viewed\_categories add Current Page Content’s category”. Bubble handles lists naturally.

Repeating Group recommendation search: Use “Do a search for Content” and sort by fields that matter. Bubble allows dynamic sorting using “Sort by” → “Popularity_score” or “Category intersects with Current User’s viewed_categories:count”. This creates a relevance effect.

• Filter by categories the user interacted with.
• Sort by match count or by recency.
• Add a fallback group showing generic popular content if the user has no history yet.

Example Dynamic Sort

// Repeating Group data source
Search for Content:sorted by
This Content's category is in Current User's viewed_categories: count

Optional: Using external models

You can generate smarter recommendations by sending user behavior to an API (like OpenAI or your own backend) via the API Connector. The API returns a list of content IDs, and then you display those items using “Do a search for Content where unique id is in API’s response”. This is fully supported by Bubble.

Practical Tips

• Always store interactions immediately using “Make changes to User” or “Create a new Thing”.
• Use Privacy Rules so users can only see public content.
• Use backend workflows if you need nightly recalculation of popularity scores.