Knowledge Similarity

Available in Lightning Experience. View supported editions.

Duplicate and near-duplicate articles make it harder to find the right answer and multiply the content that you maintain. Knowledge Similarity helps your authors catch overlap early before another duplicate article is published.

When an author checks for similar articles, Knowledge Similarity compares the content in the article against your existing articles and returns the closest matches, each with a similarity rating. Knowledge Similarity ranks matches as Very Low, Low, Medium, High, or Very High, based on how closely the content aligns with the article the author is writing or viewing. Authors decide whether to open a match and improve it, or continue with a new article.

Knowledge Similarity is grounded on an Agentforce Data Library in Data 360, which indexes your knowledge articles. When an author checks for similar articles, the feature retrieves the most similar articles from that indexed content and ranks the matches. To get the most accurate matches, ground Knowledge Similarity on a data library that has an enriched search index.

Knowledge Similarity uses semantic matching to identify similar articles based on meaning and intent, instead of exact keywords. For example, it can connect search queries such as “Wi-Fi not working” with an article titled "Resolving wireless connectivity issues." Knowledge Similarity identifies near-duplicate articles and surfaces relevant content even when the phrasing differs.

• Knowledge Similarity Considerations
  Review these considerations before you set up and use Knowledge Similarity.
• Set Up Knowledge Similarity
  Set up Data 360 and an Agentforce Data Library, then turn on Knowledge Similarity so your authors can check for similar articles before they publish.
• Find Similar Knowledge Articles
  Knowledge Similarity compares a draft or published article against your knowledge base and returns a ranked list of similar articles. Use Knowledge Similarity before publishing new content to avoid duplicates, or when reviewing existing articles to identify coverage gaps.