Research3 — AI Research Intelligence Platform
An AI search engine for scientific research that helps people find peer-reviewed work, understand complex papers and turn findings into insights they can verify.
Research3 is an AI-powered platform for scientific research, built for researchers, academics, students and knowledge workers. It brings scientific search across more than 100 million peer-reviewed papers, document upload and chat, cited AI answers, summaries and explanations, and research-integrity auditing into one workflow. Research is already complex. Adding AI creates a second problem: trust. The product has to make dense technical information easier to understand without turning the AI into a black box that nobody can verify.
Client
M0
Services
Product Design UX Strategy Landing Page Design Visual Direction
Industries
AI & Machine Learning, Scientific Research, Research Intelligence
Date
Jan 2026
How do you make AI useful for research when every answer has to hold up to scrutiny? As Design Lead at M0, I worked on Research3’s landing-page structure, product UX and visual direction, shaping how a researcher moves from a question to evidence they can trust. On the landing page, I structured the story as a researcher’s path: finding papers, understanding them, exploring new directions and verifying the work. Design Challenges & Solutions 1. Dense Information, Limited Attention Challenge: Researchers face huge amounts of dense information, and showing everything at once creates overload. Solution: Built a clear hierarchy that moves from search to paper to relevant insight to deeper analysis, where each layer answers one question before offering the next. This lowers cognitive load and lets users go deeper step by step. 2. Answers You Can Check Challenge: AI-generated research answers are hard to trust when users cannot see where they come from. Solution: Designed citation-led interactions in which every answer carries its document references, with a clear visual difference between the AI’s interpretation and the underlying evidence, so users verify information instead of accepting it blindly. 3. Four Tasks, One Journey Challenge: Search, document chat, summarization and auditing are different mental models. Solution: Designed separate but connected workflows for discovery, document analysis and research validation, giving each task a clear purpose while keeping the research journey continuous.
Outcome Research3 is live at research3.ai as an AI search engine for scientific research, bringing verified research, cited answers from your own documents and AI-powered research auditing into a single workflow. Every insight traces back to its original source. Key Learnings Trust Is a Design Material Learning: In AI products for research, confidence comes from evidence, not from fluent answers. Implementation: Citations and references were treated as primary interface elements, not footnotes. Depth on Demand Learning: Expert users want depth, but not all at once. Implementation: Progressive layers let a researcher scan first and open the detail only when it matters. Separate Modes, Shared Context Learning: Different tasks need their own space, but switching between them should not break a researcher’s train of thought. Implementation: Discovery, analysis and validation were designed as connected steps of one journey. This project sharpened how I design for high-stakes information: speed matters, but never more than the ability to check the work.
