Evaluating Web of Science’s AI-Powered Research Assistant

Can an AI tool improve research outcomes?

Posted on June 4, 2025 in Blog Posts

Core to the mission of Choice Reviews is providing succinct and helpful reviews of the newest academic resources to empower librarians to make the best selections for their collections. Over the years, we have expanded our understanding of academic resources from scholarly books to include electronic resources and database subscriptions. More recently, we have added to our offerings by reviewing AI tools. Librarians are often responsible for subscribing to or advocating for these tools and services, so we want to equip them to make smart decisions about what to purchase and why. Especially where AI is concerned, we want librarians to be able to “separate the tools from the toys,” to put it in Choice Editor and Publisher Rachel Hendrick’s words.

In our upcoming July 2025 issue of Choice, we are running Jason A. Reuscher’s review of Web of Science‘s AI-powered Research Assistant. Reuscher is a Research and Instructional Librarian at Pennsylvania State University. Reuscher’s review of the Research Assistant, posted below, details the many dimensions of this tool and evaluates its usefulness to academic researchers.

---

🔥 Stay up-to-date with LibTech Insights by signing up for our free newsletter. Just one weekly email with our new blog posts, top tech news stories, and other bonus content. Check out some posts from our archive:

• Key Library Tech Takeaways from the 2025 ACRL Environmental Scan
• Enhancing Reference Interactions with Scopus AI
• 5 Findings to Know about Your Gen Z Patrons’ Tech Habits

---

Review of Web of Science Research Assistant by Jason A. Reuscher

Web of Science (WoS) has released a Research Assistant (RA) intended for research and scholarship at all levels, from undergraduate to faculty and professional study. Because the WoS Core Collection comprises 120 years of indexed research, RA is a new way to visualize and interact with the scholarly record, leaving the detritus of the web outside of its search parameters. As a premium service, it has an additional cost to the existing WoS Core Collection.

Multilingual in construct, the AI large language model is designed from ChatGPT but draws from the WoS Core Collection using natural language processing to search efficiently. RA is integrated into the WoS platform next to the traditional search features, and it is easy to find and use when subscribed. Clicking on it allows you to start a chat by asking a question or by building your research skills with one of three guided tasks: understanding a topic, creating a literature review, and finding a journal to publish your research. There are also example questions, which can help users understand current research topics and reveal the types of questions that RA can successfully parse.

Results from any of the features are displayed as a chat-style thread. There is an option to see how the results were generated, which can include where the RA searched, which terms it used within a Boolean construct, and the number of results returned from that search. This is followed by the answer to the query, with linked and referenced documentation. Below the linked references, the user can continue to interact with RA, either by asking another question or by using some of the features that it recommends. These include document, data, and topic visualizations plus top authors based on citations and connectivity. RA also poses follow-up questions that are well-connected and relevant to the original query. New chats can be started at any time and are easy to initiate after an initial chat, and there is a chat history that logs searched topics with the option to delete chats anytime.

Some topics seem more developed than others based on what the WoS Core Collection holds. If the RA is faced with an unanswerable question, it provides a best guess using its relevancy algorithm with citations included. When posed the same topic twice in the guided task of understanding a topic, there was some small variance in RA’s answers but nothing of concern. When asked a similar question, however, it did not provide enough variance in its response from the original questions. Additionally, when using the Literature Review guided task, there was an internal error that occurred while attempting compilation of the literature.

Research Assistant is in its infancy, yet it is a much better platform for research inquiries than the existing free models because it is limited to use only vetted research. It is worth investigating, especially if your institution subscribes to the existing WoS Core Collection.

Summing Up: Recommended. Lower-division undergraduates, upper-division undergraduates, graduate students, and researchers.

---

⭐️ Found this review helpful? Learn more about subscribing to Choice Reviews at your library for up-to-date reviews of AI products, academic literature, and much more.

Related Posts