Best practices to improve Catalog Item discoverability (Suggested Results & Search Results)
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
an hour ago
Hi everyone,
I'm looking for best practices to improve the discoverability of Catalog Items in AI Search, both for:
- Suggested Results
- Search Results
What I observed
- Enriching the Short Description improves the Search Results, but has no impact on the Suggested Results.
- Adding the Meta field to the OOTB Catalog Item Indexed Source, then reindexing, also improves the Search Results, but still has no impact on the Suggested Results.
- Suggested Results also behave differently depending on the environment and even between users who satisfy the same User Criteria.
According to the ServiceNow documentation, Suggested Results use a dedicated relevancy model based mainly on:
- Title matching
- Freshness
Another point that surprised me is that the Semantic Search Mappings only contain:
- name → title
- name → body
- description → body
short_description is not mapped, yet updating it clearly improves the Search Results.
Questions
- What are the recommended best practices to improve Catalog Item discoverability (both in Suggested Results & Search Results) ?
- Is adding the Meta field to the OOTB Catalog Item Indexed Source considered a supported/recommended customization?
- Could this have unintended impacts on other consumers such as Now Assist?
- Is using Meta preferable to creating Relevancy Improvement Rules?
- Do Relevancy Improvement Rules affect only Search Results, or also Search-based Suggested Results?
- Is there any supported way to improve Suggested Results beyond optimizing the Name/Title?
- Is it expected that short_description improves Search Results even though it is not included in the Semantic Search Mappings, but is configured in the Field Settings & Mapping (map_to = text)? How is this field used by AI Search?
- Why could Suggested Results behave differently between environments, or even between users, when the users satisfy the same User Criteria?
- Is there any way to explain or debug the ranking of Search Results and Suggested Results (for example, understanding why a Knowledge Article is ranked above a Catalog Item)?
Any feedback or experience would be greatly appreciated.
Thanks!
- Labels:
-
AI Search
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
23m ago
Hi @Jonathan Rousse ,
Based on what you've described, the behavior seems expected
Item Name/Title has the biggest impact on Suggested Results, while Short Description, Meta keywords, and Synonyms tend to have a stronger influence on Search Results.
Adding the Meta field to the Catalog Item Indexed Source is generally a supported AI Search configuration and is commonly used to improve search relevance.
Meta is usually a better long-term approach than relying heavily on Result Improvement Rules, which are best reserved for specific ranking exceptions or business-driven boosts.
Result Improvement Rules primarily affect Search Results. They do not appear to have a significant impact on Suggested Results.
The effect of `short_description` is expected because fields mapped as `text` are still indexed and used for retrieval, even if they are not explicitly listed in Semantic Search Mappings.
Differences in Suggested Results between environments or users can be caused by indexing status, search analytics, content freshness, search profiles, and AI Search relevance signals, even when User Criteria are the same.
To understand ranking differences, compare the indexed content, synonyms, search profile configuration, and any improvement rules. A Knowledge Article may rank above a Catalog Item because of stronger title matching, richer content, freshness, or higher engagement signals.
My recommendation would be to focus on Title, Short Description, Meta keywords, and Synonym Dictionaries first, and use Result Improvement Rules only when specific searches need manual tuning.
