Why trusted content matters more than ever
As AI tools become increasingly integrated into research workflows, questions of trust become more vital
From Elsevier
The introduction of AI tools presents a new potential for misinformation because of their ability to hallucinate responses, including citations (Westreich, 2025). False citations and hallucinations stemming from GenAI tools waste researchers’ time and increase the likelihood of retractions in the published record. At worst, these tools expose researchers or their institution to reputational and financial costs, put grant funding at risk and degrade trust in the research ecosystem and higher education.
Despite these concerns about the trustworthiness of AI tools, almost 60% of researchers use AI tools for work and 69% expect that AI tools will save them time in the next several years (Researcher of the Future — a Confidence in Research Report, 2025).
In response to these developments, the ability to discern trusted research and citations has never been more critical, particularly within AI tools. To mitigate risks and enhance benefits, AI tools used in research should be built on trusted content with appropriate safeguards and supported by responsible use and AI literacy.
Content for research when using AI tools
For AI tools to support research effectively, the quality of the content must meet high standards of research excellence and integrity. Research-grade AI tools are grounded in curated, peer-reviewed literature which is developed in line with established standards of research integrity. This means the content within these tools reflects the core principles of accuracy, authority and transparency, reinforced through processes like peer review and ongoing integrity oversight.
In contrast, general-purpose AI tools typically draw on open web content with varying levels of quality, transparency and oversight. While useful for general tasks, they often lack consistent standards for evaluating sources and present information with equal confidence regardless of its credibility. This can introduce risks such as bias, outdated information or unclear provenance.
Curating tools for trusted content
Ensuring access to tools built on trusted content is an extension of the role librarians play in curating and evaluating scholarly resources. Similar to assessing journals and databases, librarians can apply the same principles to AI tools: examining underlying data sources, transparency and alignment with scholarly standards alongside functionality.
In doing so, they also continue their role as educators within the research process. Supporting researchers in understanding different tools, encouraging critical evaluation of AI-generated responses and advocating for systems that prioritize integrity are all familiar aspects of information literacy, now applied within an AI context.
Evaluating research-grade AI tools
Not all research-focused AI tools are equal. Librarians and their institutions can use the categories below to consider different tools.
Content quality and rigour
Types of content: A quality research-grade AI tool is typically built on a trusted foundation of scientific research: peer-reviewed literature. Some tools built for research purposes include pre-prints, conference proceedings, reviews and other material that hasn’t been peer-reviewed. These materials can support some research goals, but it’s important to evaluate their quality and role. Either way, ensure the AI tool responses contain citations to verify sources.
Content curation: In addition to content type, it’s important to understand the tool’s content selection process or in other words: who or what determines which content is included – and excluded – from the tool? Human-curated content strategies with a set framework for inclusion can help to strengthen quality and credibility. Using independent review boards to select content adds another layer of trust and rigor to the results by increasing transparency, authority and providing consistent oversight.
Text availability
Within tools using peer-reviewed content, coverage can vary widely, including abstracts, full text journal articles and book chapters. The best coverage depends on the research goals and stage. A wide array of abstracts works well for gaining breadth of current research, which is better suited for beginning a research project.
Full text has two main types: journal articles and book chapters. Full text journal articles provide the latest updates and research in a field, while book chapters provide in-depth exploration and a cumulation of knowledge on a specific topic. In contrast to abstracts, full text has an in-depth view of the content, allowing for investigation into the findings, methodologies and limitations. This is particularly important when synthesising relevant articles and identifying gaps for new research.
Volume
This refers to the size of the content base within the AI tool, specifically full text access as abstracts are available to everyone. Tools using peer-reviewed content can include open access as well as subscription content. Greater volume of content can reduce the chance of missing an article relevant to the research topic. When full-text coverage is limited, confidence in the completeness of a response may also be reduced.
Subject
Each content base has a different mix of subject makeup, depending on where they receive their content. As collaborative and interdisciplinary research are on the rise, tools with a broader range of subject coverage can meet the needs of a changing research ecosystem (Researcher of the Future — a Confidence in Research Report, 2025).
Publisher
Understanding which publishers contribute to an AI tool’s content base supports understanding the quality of the content included. Ensuring the tool’s content comes from reputable publishers can increase confidence in the AI response as being scientifically sound. Comprehensive cross-publisher full text content and data support more informed decisions and research quality.
Update frequency
Understanding how frequently the content base is updated is essential to staying up to date on the latest research and reducing the risks of missing insights. Additionally, updates play a key role in maintaining the scholarly record. Frequent content updates help to manage retractions, provide transparency, and build confidence in the research and platform.
Trusted content in AI tools
As AI becomes more embedded in research workflows, tool choice can support more responsible use. It is essential to adopt tools built on trusted, well-governed content. In doing so, researchers and institutions can harness the benefits of AI while maintaining the rigour, reliability, and credibility.