We're reclaiming inactive PDIs to keep them available for active builders. Learn what's changing, who's affected, and how to protect your work. Read More

How Does KB Search & Recommended Actions (AI Feature) determine Article Relevance & Ranking

Jaydeep Bahulek
Giga Expert

3 weeks ago

Hello,

We would like to understand how Knowledge Base (KB) search and Recommended Actions determine which knowledge articles are returned and ranked in ServiceNow.

Specifically, I have the following questions:

  1. What criteria are used by KB Search and Recommended Actions when searching for relevant knowledge articles?
  2. Does the search algorithm consider the KB article affected version (for example, version numbers or article revisions) when determining relevance?
  3. If multiple versions of the same article exist, how does ServiceNow decide which version to surface in search results or recommendations?
  4. What factors influence the ranking of results (e.g., title match, keywords, article content, metadata, popularity, recency, user behavior, AI Search relevance, etc.)?
  5. Are there any differences in how standard Knowledge Search and Recommended Actions identify and rank articles?

Any insights into the underlying search logic, indexing process, or best practices for improving KB discoverability would be greatly appreciated.

Thank You!
JD

0 Helpfuls
  • 265 Views
1 REPLY 1

1__PrabakaranC
Tera Expert

3 weeks ago

As far as I know, KB Search mainly looks at the article title, content, short description, and keywords when finding relevant articles. If AI Search is enabled, it can also use semantic matching and other relevance signals.

I don't think the article version number itself is used for ranking. In most cases, only the latest published version of a knowledge article is indexed and returned in search results, while draft or retired versions are excluded.

For Recommended Actions, the platform also takes the context of the record (such as the incident short description and details) into account when suggesting articles, so the results can differ from a standard knowledge search.

I'd be interested to know if anyone has seen documentation confirming whether article revisions or version history play any role in the ranking logic.

*************************************************************************************************************
Thanks,
Prabakaran C
LinkedIn: https://www.linkedin.com/in/prabakaran-c-534b6826a
****************************************************************************************************************

0 Helpfuls