Skip to main content

​
What you can do

  • Search your blog articles using natural language queries.
  • Adjust ranking weight sliders to control how much semantic similarity and freshness influence results.
  • View per-article score breakdowns showing how each signal contributed to the ranking.

​
Run a content search

  1. Go to Evaluate → Content Search.
  2. Enter a search query in the search bar (e.g., “sustainable fashion tips”).
  3. Review the article results, which are ordered by their combined ranking score.
Each result card shows the article title, author, publication date, featured image, and summary.

​
Tune ranking weights

The content search workbench includes sliders for two signal groups that control how articles are ranked:
  • Semantic — How closely the article’s text and images match the meaning of your query. Higher values favor articles with content that is semantically relevant to the search terms.
  • Freshness — How recently the article was published. Higher values push newer articles toward the top of results.
To adjust weights:
  1. Expand the Content Search Workbench panel.
  2. Drag the sliders to increase or decrease each signal group’s contribution.
  3. The weights are automatically normalized so they always sum to 100%.
  4. Results update immediately to reflect your changes.

​
View score breakdowns

When you run a search with the workbench open, each article card displays a score breakdown showing:
  • The overall ranking score.
  • The raw and weighted contribution from each signal group (semantic and freshness).
This helps you understand why a specific article ranks where it does and how weight adjustments affect ordering.

​
Tips

  • If results feel too focused on older evergreen content, increase the freshness weight to surface recent articles.
  • If recent but less relevant articles rank too high, increase the semantic weight to prioritize topic relevance.
  • Use the evaluate page to preview how weight changes affect results before applying them to your live content search via the API’s tuning.rankingWeightOverrides parameter.

​
See also