Summary
EBSCO's Recommended Search Queries uses generative AI to suggest relevant follow-up searches based on a user’s original natural language query. Designed to support diverse research needs, it helps users refine or expand their searches with targeted, novel suggestions. When available in EBSCO interfaces, library administrators can enable this feature in the Configuration module of EBSCO Experience Manager.
The following post is from GenAI Recommended Search Queries – Frequently Asked Questions (EBSCO Connect article).
What problem does Recommended Searches solve for users?
Recommended Search Queries uses generative AI to suggest relevant follow-up searches based on a user’s original natural language query. Designed to support diverse research needs, it helps users refine or expand their searches with targeted, novel suggestions.
How will Recommended Searches be displayed to users?
When enabled, after conducting a Natural Language Search, users will be presented with a “You may also ask..” section beneath their search box, as depicted below:
How are the suggestions generated?
The system uses a large language model (LLM) to analyze the intent and context of your original search. Based on that analysis, it generates three follow-up queries designed to help you either dig deeper into a topic or uncover new but related directions for exploration.
Do recommended queries support filters?
Yes. If filters are applied to the original search, those filters remain active when the user selects one of the recommended search queries. Clicking a suggestion immediately runs a new search using Natural Language Search mode.
Is using recommended queries reproducible?
Yes. If a user runs the same search in Natural Language Search mode within a 30-minute window, the same recommended queries will appear. This applies whether the user repeats the search directly or navigates back to the original result list page. This caching helps ensure consistency as users move between results and other pages. After 30 minutes, new suggestions may be generated based on updated model input or data.
When will recommended queries not be displayed?
Recommended Search Queries will not appear in the following scenarios:
- When the user is searching with a search mode other than Natural Language Search mode
- When no results are returned
- When searching with a field code
- When proprietary content curation is applied to the profile
Does this work in all languages?
Currently, Recommended Search Queries are only generated in English. Introducing multilingual support adds complexity and is being evaluated for future roadmap inclusion.
Suggestions are generated in part based on the relevance and volume of results returned for a given query. As a result, searching in English may produce different recommended queries than searching for the same term or phrase in a different language or character set, due to differences in content availability in varying languages.
Can our library control whether this is available?
Yes. GenAI Recommended Search Queries are OFF by default and must be enabled by a library administrator in EBSCO Experience Manager (EEM). For those not yet set up with EEM, please see Requesting Access to EBSCO Experience Manager (EBSCO Connect article).
Click the Edit button for the Natural Language search mode in EEM to open the customizable options.
On the Edit search mode modal for that search mode, check the box for the new “Enable suggested search queries” setting. When it is checked, the suggested searches feature will be enabled on the user interface.
Please note that the Natural Language Search mode Visibility setting must be toggled to Visible for the recommended searches to show on the user interface.
How is the quality of GenAI Recommended Search Queries ensured?
EBSCO employs a rigorous quality assessment process, including Human-in-the-Loop (HITL) reviews, internal testing by search SMEs, and user testing with students, academic researchers, librarians and clinicians. This quality assessment helps refine the AI's output, improve accuracy, and protect against biases in the AI’s recommended queries.
What safeguards are in place for accuracy and security?
EBSCO evaluates the AI model’s output quality and implements safeguards against adversarial prompts or unintended behavior. The goal is to ensure suggestions are relevant, secure, and consistent with academic research needs.
Are GenAI Recommended Search Queries available via API?
No, but this is in consideration for future roadmap inclusion.
Does GenAI Recommended Search Queries require a personal user account?
No, users do not need to sign in with a personal (MyEBSCO) account to see or use recommended search queries.
Does my institution need to purchase GenAI Recommended Search Queries?
No, the feature is available to institutions using the new user interface. It can be enabled in supported profiles through EBSCO Experience Manager (EEM).
How is EBSCO handling ethical concerns related to the use of AI?
Please see EBSCO Information Services Defines Guiding Principles for the Responsible Use of Artificial Intelligence (11 March, 2024) as well as EBSCO’s AI Tenets.
For any additional questions or information, please reach out to EBSCO at ai@ebsco.com.
There are multiple EBSCO resources available in the BadgerLink collection. For questions on those resources available through BadgerLink's statewide subscription, please contact us.
BadgerLink is made possible through a combination of both state funds appropriated through the biennial budget, and Library Services and Technology Act (LSTA) Grants to States funding from the Institute of Museum and Library Services (IMLS).
