Curie

Curie is an AI-powered research assistant for scientific literature and document analysis, presented as built for real scientific work rather than as a generalist chatbot. It analyzes scientific literature, processes documents and delivers intelligent insights powered by advanced AI technology. The platform adapts to many kinds of researchers — solo researchers, graduate students, systematic review teams and R&D departments — so that each can search literature, extract data and verify findings with precision. Its stated aim is to let researchers focus on the science that matters most, and you can start writing on Curie for free. Research is slowed down by repetitive work. Searching literature, screening abstracts, extracting data and drafting summaries eat up valuable hours and pull focus away from the parts of research that only a researcher can do: interpreting findings, forming hypotheses and making decisions. Curie's stated purpose is to handle those repetitive steps so that time goes back into judgment and interpretation. The problem is not only speed. Doing better research also means pushing further — covering broader literature, more complex reviews and questions that a person would otherwise not have time for. Curie is positioned as a research partner that gives researchers the confidence to take on that wider scope, so more ground is covered and more impact comes from the work. Curie's literature search does not require keywords or boolean filters. You type your question the way you would ask a colleague, and Curie figures out what to search and how to answer. Behind that plain-language interface, Curie searches PubMed, arXiv, Europe PMC, OpenAlex and Semantic Scholar in parallel, so a single question reaches multiple scientific databases at once instead of being repeated across separate tools. The site presents this focus on scientific literature as the difference from a generalist chatbot. Researchers can ask it to find recent papers on a topic, look for meta-analyses, locate studies on a specific model or method, or search PubMed and arXiv together for a subject such as quantum error correction. A central part of Curie is verification. The assistant verifies every claim against the source text, so you can see what is backed by the literature and what is not. This matters because unsupported statements are difficult to spot when synthesizing many papers. Related to this is document analysis: you can point Curie at a document and it will extract structured data, showing the quote behind each value it extracts. For example, you might extract sample sizes and outcomes from clinical trial PDFs, compare reaction yields across synthesis papers, or build a table comparing sample size and effect size across a set of studies. Because each extracted value comes with its supporting quote, results stay traceable to the original text. Curie also supports a full systematic review workflow: protocol, screening, PRISMA diagram. The important detail is that you approve every decision, so the process stays under researcher control rather than being fully automated. Researchers can draft a systematic review protocol for a topic such as antibiotic resistance studies, screen records, and produce a PRISMA diagram as part of the same flow. To keep this work organized, Curie provides Projects and a library, which hold the material and the work associated with each effort so that searches, extractions and reviews do not have to be managed across separate tools. Overall, Curie's approach combines three things that are explicitly described on its site. First, natural-language input: you ask in plain language and Curie decides what to search and how to answer. Second, parallel search: PubMed, arXiv, Europe PMC, OpenAlex and Semantic Scholar are queried at the same time. Third, verification against source text: every claim is checked against the source so backed and unbacked statements are distinguishable, and extracted data values are tied to the quote they came from. On top of that sits a human-in-the-loop model for systematic reviews, where the researcher approves each decision. Projects and a library keep the resulting work organized. The stated benefits are time and scope. Curie is presented under the promise "Save hours, not minutes," with research productivity maximized by taking over literature searching, abstract screening, data extraction and summary drafting. By removing those repetitive steps, it is intended to return focus to interpretation, hypothesis formation and decision-making. In practice that means fewer hours spent on mechanical steps every week and more capacity for the analysis that determines the quality of the output. The second benefit is reach: Curie acts as a research partner that gives researchers the confidence to tackle broader literature, more complex reviews and questions they would otherwise not have time for, meaning more ground covered and more impact from the work. The site frames this as support for both speed and ambition — working faster and pushing further. Curie's site lists concrete tasks it is used for. Researchers can summarize the latest treatments for triple-negative breast cancer; find recent papers on CRISPR gene editing for sickle cell disease; extract sample sizes and outcomes from clinical trial PDFs; explain the mechanism of action of mRNA vaccines in simple terms; search for meta-analyses on cognitive behavioral therapy for anxiety; draft a systematic review protocol for antibiotic resistance studies; compare reaction yields across synthesis papers; summarize the methodology section of an astrophysics paper; find studies on machine learning models for early Alzheimer's detection; build a table comparing sample size and effect size across studies; check whether a claim about vaccine efficacy is supported by the literature; search PubMed and arXiv for papers on quantum error correction; and generate a Python script to analyze a gene expression dataset. Questions about what current research says, such as on microplastics in drinking water, are also among the examples. Curie is presented as suitable for different kinds of researchers on one platform: academic researchers and professors, PhD and graduate students, systematic review teams, and industry R&D teams. The site shows institutions that have already tried it, including CSIC (Consejo Superior de Investigaciones Científicas), Universitat de Barcelona, Fundació Bosch i Gimpera, IrsiCaixa, Helse Bergen (Haukeland universitetssjukehus), Università degli Studi di Messina, Uganda Christian University and MICIU / AEI. Security and privacy are addressed directly: Curie states that it is strictly GDPR compliant, that its servers are located in Europe, that it does not use user data to train models or to improve Curie, and that user data is not shared with anyone. It also describes pseudonymization and anonymization of personal and identifying information in certain documents. Access is offered for free, with calls to start writing, get free access and get started. Curie's value proposition is a focused AI research assistant rather than a general chatbot: parallel search across major scientific databases, plain-language questions, verification of every claim against source text, structured data extraction with supporting quotes, and a supervised systematic review workflow with protocol, screening and PRISMA diagram. Combined with projects, a library and a privacy posture aimed at sensitive research data, it is built to save researchers hours each week and to widen the scope of questions they can take on.