Indexing is the foundation that powers Supercharger’s AI features, especially semantic recommendations. This page explains what it is, why it’s needed, and how it works in plain language.
What Is Indexing?
When Supercharger “indexes” a post, it does two things:
- Reads the post content and breaks it into manageable sections (called chunks).
- Sends each chunk to an AI provider to generate a _vector embedding_ — a numerical representation of the text’s meaning.
- Stores those embeddings in your WordPress database.
The result is that every indexed post has a mathematical “fingerprint” of its meaning, not just its keywords.
Why Does This Matter?
Traditional related-posts features work by matching tags, categories, or shared keywords. If two articles cover the same topic using different words, they won’t be matched.
Vector embeddings solve this. Because they capture _meaning_, Supercharger can find articles that are genuinely related even if they don’t share a single keyword. A post about “managing anxiety” and one about “dealing with work stress” would be recognized as semantically similar.
This powers all modules that recommend or recirculate related posts: AI Recommendations, AI Inline Recommendations, AI Continue Reading Chips, AI Footer Recirculation, AI Smart Up-Next, and AI Exit-Intent Recirculation.
How Indexing Runs
Indexing happens in the background using WordPress Cron. After you configure Supercharger and trigger indexing (either via the Setting Up or manually), the plugin:
- Builds a queue of all posts that need to be indexed.
- Processes posts in small batches on each cron run.
- Marks each post as indexed once complete.
- Re-indexes a post automatically when its content changes.
You do not need to manually trigger indexing after the initial setup — Supercharger monitors content changes and keeps the index up to date.
Monitoring Progress
The current status of your index is always visible on the Dashboard:
- Total posts in index — how many posts have been successfully indexed.
- Pending — posts queued but not yet processed.
- Failed — posts that failed to index (usually due to API errors).
You can also manually check index health from Supercharger → Tools.
OpenAI Is Required for Embeddings
Generating vector embeddings requires OpenAI specifically (model: text-embedding-3-large). This is a current requirement regardless of which provider you use for text generation.
This means:
- You can use Anthropic or DeepSeek for generating summaries and highlights.
- You still need an OpenAI API key if you want semantic recommendations to work.
If you don’t have an OpenAI key, you can still use all content-enhancement modules (Paragraph Highlights, Quote Puller, Content Summarizer, Key Moments, Key Questions, and Excerpts). All modules that find and recommend related posts will be unavailable.
API Usage During Indexing
Indexing does consume API credits. Each post generates one or more API calls depending on its length. Longer posts with more content sections will use more tokens.
To manage costs:
- Start by indexing only the post types that matter most (e.g., Posts only, not Pages).
- Monitor your usage in your AI provider’s dashboard.
- Use the index tools to avoid re-indexing content unnecessarily.
Re-indexing
If you change your AI provider or embeddings model, your existing embeddings will no longer be compatible. In this case, you’ll need to run a full re-index from Supercharger → Tools.