AI-powered searching for novices

Source: Hierarchical AI Graphic from Preisler, 2024, pg.6. Compiled by Dean Giustini, UBC Biomed librarian, dean.giustini@ubc.ca Updated 4 April 2026 | Part of Knowledge Synthesis (KS) & AI Search Wiki 2026 & A to Z Listing See also Artificial intelligence (AI) glossary of terms | AI-powered searching - a definition Note: OpenAI (and thus ChatGPT) leads the general AI space, but AI companies such as Claude by Anthropic and Google Gemini are developing deep research tools and others such as Elicit.com and Perplexity are experimenting with AI-powered academic searching in support of research. Which companies are behind AI search tools? How will AI tools affect our traditional bibliographic databases? Will we see GenAI being put into our search platforms? Can we stop it? Algorithms | Which companies are behind AI search tools? | Vector-based searching and embeddings Large language models (LLMs) | Machine learning | Neural networks | Prompt engineering | Reasoning models AI-powered search tools: Asta (agentic AI by Ai2), Consensus, Elicit.com, Open Evidence, otto-SR, Perplexity, PubMed.ai, Undermind.ai Also: This open textbook (or wiki channel) is intended to help librarians and other information professionals learn about AI. It is not, in itself, meant to be seen as promotion of AI. If anything, the goal is harms mitigation or harms reduction. Definition "AI-powered searching" (or AI-enabled searching) refers to systems with AI features and/or supported by machine learning and/or natural language processing techniques. The goal is to enhance the location, collection and synthesis of papers but beyond traditional searching techniques. Other synonyms for this type of searching include AI-powered discovery systems or AI search and synthesis tools. 1. What AI-Powered Searching Is AI-powered searching refers to the emergence of web-based search tools that use large language models (LLMs) and generative AI to search and synthesize the literature. Unlike traditional keyword searching, AI-powered systems use artificial intelligence (AI) — including natural language processing, semantic understanding, and machine learning — to interpret queries and retrieve contextually relevant research. These tools go beyond bibliographic database searching by extracting insights and interpreting meaning, rather than simply matching words. Key Beginner Takeaways Re: AI-Powered Searching AI tools interpret your natural language and controlled vocabulary queries AI tools summarize and extract data from research papers AI tools identify relevant literature more quickly than traditional methods but miss papers found in comprehensive searching 2. Core AI Concepts That Matter for Search Before using AI-powered search tools, novices should understand the basic AI concepts that underpin them. Essential Concepts Large language models (LLMs): The engines behind many AI search assistants Semantic search and vector embeddings: Techniques that allow AI systems to understand meaning and similarity, not just keywords Prompt engineering: How questions are framed to improve search results These concepts appear throughout the Wiki and are essential for understanding how and why AI-powered search tools behave as they do. 3. Landscape of AI-Powered Search Tools Much of this Wiki focuses on practical AI-powered tools, many of which support literature discovery and knowledge synthesis. Common Examples Elicit.com: Automates literature searching and summarization Perplexity, PubMed.ai, Scite.ai: Tools with different strengths in summarization, biomedical search, and citation analysis Open Evidence, otto-SR, Undermind.ai: Tools that support deeper or more automated review workflows For beginners, recognizing that different tools have different strengths — and operate in different ways — is an important early insight. 4. Ethics, Trust, and Critical Evaluation The Wiki emphasizes that AI-powered search tools are not inherently reliable sources of information but they can be integrated into early searching, during testing phases, and to translate completed human-derived searches. They may generate outputs that appear plausible but are often incorrect, lack transparency, or reflect biases in training data or design. Beginner Responsibilities Question the credibility and transparency of outputs Understand ethical constraints in research and knowledge synthesis Validate AI-generated results rather than accepting them at face value These considerations are essential for responsible and scholarly use of AI-powered search tools. 5. Positioning AI Within Traditional Knowledge Synthesis AI-powered search tools are not replacements for established knowledge synthesis methods. Instead, they should be used to augment traditional approaches. Established Methods Include Systematic reviews Scoping reviews Meta-analysis Understanding where AI fits within these workflows — and where human expertise remains essential — is critical for novices applying AI tools in academic, clinical, or policy settings. Beginner Learning Path for AI-Powered Search in Knowledge Synthesis This section provides a structured, novice-friendly learning path for using AI-powered search tools in knowledge synthesis (KS). It is intended for students, clinicians, researchers, and librarians new to AI-assisted searching and evidence discovery. AI-powered search tools can significantly accelerate early-stage research tasks, but they must be used critically and in conjunction with established knowledge synthesis methods. Who This Page Is For Beginners to AI-powered searching Graduate students and trainees Researchers starting scoping or systematic reviews Librarians and information professionals introducing AI tools Learning Objectives After completing this simple pathway, users should be able to: Understand what AI-powered search is (and is not) Explain key AI concepts relevant to search Identify major categories of AI search tools Critically evaluate AI-generated outputs Position AI appropriately within KS workflows Step 1: Understanding AI-Powered Search AI-powered search differs from traditional keyword searching by using artificial intelligence techniques such as natural language processing and semantic analysis to retrieve conceptually relevant information. Key ideas for novices: Queries can be written in natural language Results are based on meaning, not exact keywords Outputs may include summaries, extracted data, or ranked evidence Important: AI-powered search tools support discovery and sense-making, but they do not replace rigorous search strategies. Step 2: Core AI Concepts for Beginners Novices do not need technical expertise, but should understand the following foundational concepts: Large Language Models (LLMs): Generate and summarize text based on patterns in training data Semantic search: Retrieves results based on conceptual similarity Prompting: How questions are framed affects outputs Hallucinations: AI systems may produce confident but incorrect information Rule of thumb: Always verify AI outputs against original sources. Step 3: Understanding the AI Search Tool Landscape AI-powered search tools vary widely in purpose and design. Beginners should focus on understanding tool categories rather than mastering individual platforms. Common categories include: Literature discovery and scoping tools AI answer engines with citations Citation context and evidence mapping tools Domain-specific search tools (e.g., biomedical or policy) Beginners are encouraged to: Start with one or two tools only Compare AI results with traditional databases (e.g., PubMed) Examine how citations are generated and displayed Step 4: Critical Appraisal and Ethical Use AI-powered search tools introduce new risks that require critical evaluation. Key issues to consider: Incomplete or biased coverage Non-transparent ranking algorithms Fabricated or inaccurate citations Over-reliance on summaries Ethical use requires: Transparency about AI use Verification of all cited sources Awareness of equity and bias concerns Step 5: Positioning AI Within Knowledge Synthesis AI-powered search tools are best used to support — not replace — established KS methodologies. Appropriate uses include: Refining research questions Conducting preliminary or scoping searches Identifying themes and key concepts Supporting screening prioritization AI tools should not replace: Comprehensive systematic searches Dual screening and data extraction Risk-of-bias assessment Final evidence synthesis decisions Key Takeaways for Novices AI search is fastest and safest early in the review process AI outputs require human judgment and verification Traditional databases remain essential Critical appraisal skills are more important than tool choice Retrieval Models: BM25 and Vector Embeddings AI-powered search tools rely on underlying retrieval models to identify relevant documents. BM25: A traditional keyword-based ranking algorithm that retrieves documents using exact term matches. BM25 is widely used in bibliographic databases and supports transparent and reproducible searching. Vector embeddings: Semantic representations of text that allow AI systems to retrieve documents based on conceptual similarity rather than exact wording. Many modern AI-powered search tools use a hybrid approach that combines BM25 with vector-based retrieval to balance precision and recall. See Also Knowledge Synthesis (KS) & AI Search Wiki 2026 AI-powered searching - a definition Systematic reviews Scoping reviews Disclaimer Note: Please use your critical reading skills while reading entries. No warranties, implied or actual, are granted for any health or medical search or AI information obtained while using these pages. Check with your librarian for more contextual, accurate information.