Smart Search in Web of Science provides a faster and more intuitive way to discover research, and will soon become the default experience for all users.
As search technology evolves, users expect systems to understand context and intent without requiring technical input. Against this backdrop, we developed Smart Search, a new entry point to Web of Science content designed to make research discovery easier and more intuitive. We are excited to announce that Smart Search will soon be the default experience for all Web of Science users.
Aligning with user expectations
We built Smart Search based on direct feedback from our global research community. Surveyed Web of Science users have increasingly expressed a desire for a more intelligent search experience. Many users find traditional search interfaces too rigid, preferring systems that interpret natural language and deliver relevant results without complex syntax.
Smart Search is designed to meet these expectations. It understands user intent and combines that with updated retrieval methods to create a more forgiving search experience. Key features include:
- Entity recognition: Enter descriptive phrases and queries and get results that match your intended meaning.
- Vector and Boolean search: Retrieve results based on both semantic similarity and keyword matching.
- Multi-language support: Search in your preferred language and view translated abstracts and titles.
- Transparent results: Toggle between Boolean, Semantic or Combined results to suit your preferences.
Tailoring search to academic content
Smart Search utilizes a Natural Language Processing (NLP) parser, specifically designed to comprehend the language of scholarly research. It uses custom-built Named Entity Recognition (NER) models trained on academic content to identify key elements such as authors, journals, institutions, and scientific keywords. These models are fine-tuned to handle complex terms and multi-word phrases common in research, resulting in a search experience that is fast, precise, and tailored for academic discovery.
Incorporating semantic search and translation
Smart Search relies on established Artificial Intelligence (AI) technologies that are foundational to modern search design, helping users find relevant content more efficiently. One example is embedding-based vector search, a technique that improves relevance by interpreting the meaning behind a query rather than relying solely on exact keyword matches. These embeddings are indexed in our search engine, enabling queries to return relevant content based on similarity scoring. This approach helps users surface content that may be phrased differently but is conceptually aligned—an especially useful feature in interdisciplinary research. Smart Search blends vector search with traditional Boolean logic. Users benefit from the transparency and reproducibility of Boolean search, while also ensuring that they don’t miss important content that may be phrased differently. Users can switch between results sets based on the search method used to derive them.
Smart Search also ensures that language barriers don’t impede discovery. To support global researchers, it uses Large Language Models (LLMs) for translation—activating them only when a non-English query is detected. The LLM performs a direct translation into English, preserving the original query’s meaning and structure. From there, our custom NLP parser takes over, ensuring the query is processed with the same precision as a native English input. On the results page, translation is user-driven. Researchers can choose to view abstracts and titles in their preferred language by toggling translation on.
Implementing responsible AI to enhance search
It is important to note that Smart Search makes limited use of generative AI and does not employ autonomous AI agents. These technologies offer transformative potential, but their adoption requires careful consideration first. We understand that many libraries are still in the process of evaluating new technologies and developing guidelines for their institutions. To align, we have focused on foundational AI techniques to modernize the general search experience for a broad set of institutions and users.
We are committed to developing responsible AI, and data privacy and trust are top priorities when designing our tools. We do not train public LLMs, nor do we share publisher content, library-owned materials or user data with them for any purpose. As we continue to innovate, our focus remains on building AI tools that are not only powerful and efficient but also aligned with the values and standards of the research community.
Community feedback: listening, learning and evolving
Smart Search was first tested during a limited beta phase with development partner institutions, who had site-wide access for up to 12 months. Their feedback helped shape the experience before it was released as an opt-in feature in April 2025 and later transitioned to opt-out in July.
We’ve continuously monitored user engagement and feedback throughout the rollout, using those insights to refine the experience and guide the timeline. Between the April and July releases this year, we saw strong engagement, with the majority of users giving Smart Search a positive rating.
Since the July release, 97% of unique visitors have opted to keep it as their default.
Supporting diverse needs
We understand that search preferences vary widely. These preferences can develop over a career and shift from task to task, for example, exploring a new topic versus conducting a rigorous systematic review. This is why Smart Search complements but does not replace existing functionality.
Web of Science now offers three distinct search experiences: Smart Search, Advanced Search and the generative and agentic AI-powered Web of Science Research Assistant. Smart Search is ideal for quick, intuitive discovery using plain language. Advanced Search remains essential for users who need precision and reproducibility through Boolean logic and fielded queries. For those exploring unfamiliar topics or synthesizing literature, our research assistant provides a conversational, guided experience through common research tasks. Together, these options ensure that researchers of all skill levels can utilize the most suitable tools for their search journey.
Preparing for the next step
On November 20, 2025, Smart Search will become the default landing experience for all users. The opt-out toggle option will be removed from the interface, but Advanced Search will remain fully accessible for those who prefer fielded, structured or cited reference searches.
Respecting user preferences is central to our design philosophy. Individual users can adjust their default landing page when logged in, and librarian administrators can set a default search page for their institution to ensure that the most appropriate experience is easily accessible.
We are committed to supporting all users through this transition and continuing to refine Smart Search based on community feedback. Smart Search is designed to help students and seasoned researchers alike discover research more effectively. The evolution of the Web of Science search experience reflects our dedication to building tools that not only meet today’s needs but also anticipate tomorrow’s challenges—empowering the global research community to move knowledge forward with confidence and clarity.
Related insights
The latest news, technologies, and resources from our team.
