Responsible Literature Searching Guide

Topics

• Database Selection
• Literature Search String Development
• Identifying and Evaluating High-Quality Evidence

Limitations & Questions

• What is the corpus of the tool? Do they exclusively search open-access content? Are they transparent about which sources they draw information from—web resources, published articles, preprints, etc.? Many AI tools search Semantic Scholar or OpenAlex, which have wide coverage, but do not offer the same breadth of coverage as the library's bibliographic databases.
• Recognize that no database or AI tool contains access to ALL literature. Best practice is to search more than one database or tool to cast a wide net. AI may be biased towards certain types of literature, such as the most accessible.
• Do not become overdependent on AI. Apply your creativity and critical thinking to the development of searches.
• Be careful about uploadin full-text articles into an AI tool. You may violate copyright or license agreements. For example, do not upload a copyrighted article as a "seed article" in a literature mapping/network analysis tool.

Why These, Not Those?

• The library's bibliographic databases are trusted resources. The content goes through a formal vetting process, and can be trusted more than other AI, which may search general web content, be less controlled, and be less transparent about their corpora and processes. The publications indexed are listed within the database and the vetting processes are available on the vendor website.
• Many external AI tools use Semantic Scholar or OpenAlex, alongside agreements with select publishers, as their corpora, but they do not offer the same breadth of subject coverage as the library's bibliographic databases.
• Library bibliographic databases are already equipped with AI-enhanced features and continue to make improvements to integrate AI further, thereby enhancing accessibility and furthering research capabilities.
• You can receive more support for the use of these tools through consultation with a librarian or vendor support lines. Librarians are expert searchers and can advise you on how to search effectively.

Note: All AI can hallucinate, provide poor quality outputs, or oversimplify complex topics. Apply critical analysis to any output.

Use the ROBOT test to evaluate AI tools.

Published by the American Chemical Society, SciFinder can provide access to literature in chemistry and related fields.

• SearchSense – An AI-powered single "omnibox" search, enables natural language queries across all content. It delivers instant answers and highlights research trends through reference analytics like publication patterns and top concepts. Starting in early October, AI-generated summaries from factual data will help users quickly grasp research results and trends.

CINAHL is a database of nursing and allied health journals. There are several AI-powered search options. You can change your search technique under "Search options" in the Advanced Search interface. You can also toggle the Natural language search on and off from the database main search screen.

• Proximity searching – Searches for your terms near each other, default for the UTSW CINAHL interface.
• Natural language searching – Allows you search in a more natural, conversational manner. Select "Show refine query" to see how the system translated your query into a Boolean search.
• SmartText searching – Natural language searches, summarizing your main words/phrases, generating a corresponding search, and then identifying relevant results. You can upload larger bodies of text.
• PICOT searching – Enter your PICOT elements and the system translates your question into a keyword search with Boolean operators.
• AI Insight Summary – It is an LLM-generated 2-5 bullet point summary of the article, allowing you to quickly determine if an article is relevant to your research. The option appears when the appropriate full text is available and the content is long enough to allow generation.

Many of these same options appear in the other EBSCO Databases.

The Library provides access to several databases on the Ovid platform, including MEDLINE (biomedical), Embase (biomedical/pharmacological), PsycInfo (psychological), and HAPI (measurement instruments). AI is incorporated in a few different ways.

• Automatic Term Mapping – The database defaults to map to the closest corresponding subject heading, or you can use the Term Finder tool to identify relevant vocabulary.
• AI Article Summary – A bulleted summary of an article, including methodology, results, limitations, and clinical implications. This feature is currently only available on HTML full-text articles from proprietary Lippincott journals on Ovid.

PubMed is the National Library of Medicine's Database of biomedical literature.

• Automatic Term Mapping – When you enter a general search, PubMed will automatically map to the closest corresponding MeSH (Medical Subject Heading) term and generate a search incorporating this term and variations of your keyword(s).

Scopus is a multidisciplinary bibliographic database.

• Scopus AI – Pulls from the Scopus research corpus, specifically the abstracts, to generate summaries and provide insights on a given topic or question. It will also product concepts maps and identify subject experts and emerging themes. It will translate your query into a Boolean keyword search. Although not complete, this string can be a good starting point for a literature search. The Deep Research option will generate a multi-page report with more citations. You can view the AI's "thought" process. Scopus AI will search pull literature from 2003 onward. It is may not identify historical, seminal articles.

Web of Science is a multidisciplinary bibliographic database.

• Smart Search – Powered by semantic search and natural language processing, it allows users to enter queries in natural language and returns AI-ranked results for both documents and researchers. Boolean operators can be used but are not required. This is the default search option for Web of Science.