Home / Professions / Research & science

Best AI for research & science

Researchers at universities, universities of applied sciences, institutes and R&D departments are drowning in literature: millions of articles appear each year and a systematic review takes months. AI shifts the focus from searching and screening to assessing: semantic search finds work that keyword searches miss, screening tools surface the relevant studies, and writing aids polish English-language manuscripts. The gain lies in reproducible workflows with source attribution, not in a chatbot that invents references.

Professional tool with the strongest GDPR profileCradleGDPR: Strong
  • Provider based in the Netherlands
  • Data in the EU according to the provider
  • Publishes certification/GDPR safeguards
View the tool
General model: the default choiceClaudeGDPR: Strong
  • Editorial score 9.0 of 10
  • Elo 1506
Claude for research & science
!

Watch out in this field

Never use AI as a source: every reference must be traceable to the original article, as fabricated or wrongly attributed citations are the most common error. Unpublished data, research proposals and peer-review manuscripts are confidential or commercially sensitive, so do not enter them into a consumer account without a data processing agreement, and check whether the provider stores your input or uses it for training (publishers usually prohibit this explicitly for peer review). If you work with personal data from research participants, the GDPR and your own ethics review apply: running models locally or using an EU-hosted tool with data minimisation is then the safe route. Finally, document your search strategy and your AI use: journals and funders increasingly ask for this, and a review must remain reproducible.

All 30 options

Professional tools with the strongest published GDPR safeguards first, then general assistants by suitability.

30options, 18 field-specific
16free to try
10professional tools from NL or EU
11GDPR strong or good
Strong · 1Good · 6Limited · 8Unknown · 3
#SolutionQualityQuality from blind duels on LMArena: users choose between two anonymous answers. A difference of about 20 points is noticeable, a few points is noise. How we measurePriceGDPROur summary of what the provider itself publishes about your data. For a professional tool: whether the data is held in the EU, whether there is a certification or GDPR guarantee, and whether the provider is based in the Netherlands or the EU. For a general model: EU hosting, no training on your data and self-hosting. From Strong to Unknown; Caution only applies to a model with a documented problem. How we measure
1Cradle professional toolCradle (Amsterdam, Netherlands and Zürich, Switzerland) · NetherlandsAI platform for protein design: generates and optimises candidates for antibodies, enzymes, vaccines and peptides, and learns from lab results round by round.own modelon request (licence, no royal…Strong2ASReview LAB professional toolUniversiteit Utrecht (Netherlands) · Netherlands · free to tryDutch open-source tool from Universiteit Utrecht that speeds up the screening of titles and abstracts for systematic reviews using active learning: the model learns from your decisions and puts the most relevant studies first. Runs locally or on your own server.own modelfree (open source)Good3Keenious professional toolKeenious (Norway) · EU · free to tryNorwegian research assistant that recommends relevant articles based on your own text or topic from over 250 million papers. Works as an add-in in Word and Google Docs, with links to library systems.own modelfree plan (3 conversations per day)Good4Scholarcy professional toolEverway Opco Limited / Scholarcy (Antrim, United Kingdom) · free to tryTurns articles, book chapters and PDFs into interactive summary cards with key findings, figures, tables and references. Useful for quickly screening a large stack of literature and exporting to your reference manager.own modelfree Article Summarizer (1 same…Good5SciSpace professional toolPubGenius Inc. (Milpitas, United States) · free to tryBroad platform for literature research: searching hundreds of millions of papers, chatting with PDFs with references to the exact passage, extracting data from articles and writing with citation help.own modelfree planGood6Castor professional toolCastor (Amsterdam, Netherlands) · NetherlandsDutch eClinical platform for clinical trials (EDC, eCOA, eConsent) with Catalyst, an AI module that extracts source data from records and enters it into the study after human review.own modelon requestGood7WISKI professional toolKISTERS AG (Aachen, Germany) · EUData management platform for water and environmental measurements from KISTERS, with the KISTERS Analytics module for validation, forecasts and digital twins.own modelon requestGood8Scopus AI professional toolElsevier B.V. (Amsterdam, Netherlands) · NetherlandsGenerative AI layer on top of Scopus that answers research questions in plain language with summaries, concept maps and emerging themes, always with references to the underlying articles. Draws only on Scopus metadata and abstracts from over 7,000 publishers.on institutional licence, price on…Limited9Silvi professional toolSilvi ApS (Copenhagen, Denmark) · EU · free to tryDanish platform for systematic reviews and meta-analyses: import studies from multiple databases, screen, extract data from PDFs and visualise. Focused on transparent, traceable evidence synthesis.own modelfree BasicLimited10Consensus professional toolConsensus NLP, Inc. (San Francisco, United States) · free to trySearch engine that first retrieves peer-reviewed articles and then uses AI for a synthesised answer with citations. The Consensus Meter shows how the evidence stacks up for a yes/no research question.on free plan (10 Pro messages per month)Limited11Covidence professional toolVeritas Health Innovation Ltd (Melbourne, Australia) · free to tryWidely used platform for managing systematic reviews: importing and deduplicating, dual screening, data extraction, risk of bias and export. Part of the Cochrane review toolset.own modelSingle $339 per year (1 review)Limited12Causaly professional toolCausaly Ltd (London, United Kingdom)Agentic AI platform for life sciences that brings together biomedical literature, databases and proprietary company data in a knowledge graph and answers research questions with cited evidence. Aimed at pharma and biotech R&D.own modelon request (enterprise, via demo)Limited13Visiopharm professional toolVisiopharm A/S (Hørsholm, Denmark) · EUDanish software for AI tissue analysis in research and diagnostics, with CE-IVD apps for biomarkers in breast, lung and prostate cancer and more than 130 modular analysis apps.own modelon requestLimited14Owkin K Pro professional toolOwkin (Paris, France) · EUAI scientist for drug development: agents search for targets, predict which patients will respond and help take candidates into clinical trials, with validation in an in-house lab.own modelon requestLimited15Iris.ai professional toolIris AI AS (Bærum, Norway) · EUNorwegian AI platform that searches, summarises and extracts data from scientific literature and in-house R&D documents, including for life sciences.own modelon requestLimited16Writefull professional toolThinqLab Ltd / Writefull (London, United Kingdom) · free to tryWriting assistant specifically for academic English, with language feedback, paraphrasing, title and abstract generation and LaTeX support. Works as an add-in for Word and Overleaf and as the Writefull X web app.on free version with daily quotaUnknown17Elicit professional toolElicit Research, PBC (Covina, United States) · free to tryAI research assistant that searches literature and puts results into tables: for each article, the population, method and outcome side by side. Includes a workflow for systematic reviews with screening of thousands of papers.own modelfree BasicUnknown18ResearchRabbit professional toolLitmap Limited (Wellington, New Zealand) · free to tryVisual literature explorer: starting from one or more seed articles, it builds a network of related papers, authors and citations and learns from what you add. Free to use and widely recommended by university libraries.own modelfree (Free Forever)Unknown19Claude generalAnthropicFamily of language models from Anthropic that excels at long documents, careful reasoning, coding and reliable, safe output.1506Elo$20 / monthStrong20Gemini generalGoogleGoogle DeepMind's natively multimodal model family, deeply integrated into Workspace, with a very large context window.1527Elo€4.99 / monthStrong21ChatGPT generalOpenAIOpenAI's versatile, agentic flagship with a very large context window and a broad ecosystem of integrations.1485Elo$8 / monthStrong22Le Chat (Mistral) generalMistral AIFrench/European AI lab with open-weight and commercial models and the Le Chat assistant, with EU data residency as its core proposition.1426Elo$14.99 / monthStrong23Perplexity generalPerplexity AI · free to tryUS AI search engine that delivers answers with source citations via its own Sonar models and external frontier models, with separate Enterprise agreements on data use.–Elo$20 / monthUnknown24Copilot generalMicrosoftMicrosoft's AI layer that combines GPT and Claude models with your own organisational data and the Microsoft 365 apps.1485Elofrom €10 / monthCaution25Qwen generalAlibaba Cloud · free to tryChinese model family from Alibaba Cloud with a free chat app, a large open-weight catalogue and the proprietary flagship Qwen3.8-Max with 1 million tokens of context.1484Eloon requestCaution26GLM (Z.ai) generalZ.ai (Zhipu AI) · free to tryChinese AI lab (Zhipu, international Z.ai) with the open-weight GLM family under MIT licence, particularly strong at programming and agent tasks at low prices.1483Elo$18 / monthCaution27Kimi generalMoonshot AI · free to tryChinese AI lab (Moonshot AI) with the open-weight Kimi K2 family and an assistant that focuses strongly on agents, deep research and long documents.1487Elo$19 / monthCaution28Grok generalxAIThe model from xAI, integrated into X, distinguished by real-time access to live X posts and web data.1471Elo$10 / monthCaution29DeepSeek generalDeepSeek · free to tryOpen-weight (MIT) Chinese language model with a 1M context window and a built-in reasoning mode, at very low API costs.1475Eloon requestCaution30Llama generalMeta · free to tryMeta's natively multimodal, open-weight Mixture-of-Experts family (Scout and Maverick), free to download and self-host.–Eloon requestCaution
Download as CSV Quality: LMArena · prices: OpenRouter and providers · as of 9 November 2026

Last checked: 9 November 2026 · How we check

Where AI does the most in this field

The core tasks we track for this profession. Tools in the list above can be compared on these tasks.

  1. 1Search and explore literature with semantic search and citation networks
  2. 2Screen titles, abstracts and full texts for systematic reviews
  3. 3Extract and summarise data and study characteristics from papers
  4. 4Write and edit manuscripts, abstracts and research proposals

Every Friday: what changed in AI for research & science

New models, prices, oversight and tools, with sources and what it means for you. One email a week, unsubscribe with one click.

View a sample

We only use your address for this newsletter. You confirm first by email.

Frequently asked questions

Related professions

The full Safety grid

What is the best AI for research & science?

That depends on your requirements. Field-specific solutions such as Cradle are built for research & science; general assistants such as Claude are more widely usable and cheaper to get started with. The top of this page shows our picks.

What is the difference between a 'field-specific tool' and a 'general assistant'?

A general assistant (such as ChatGPT or Claude) is a clever all-rounder you converse with yourself. A field-specific tool is a ready-made product for research & science, often built on top of such an assistant, with its own domain knowledge, integrations and GDPR arrangements. The 'Quality' column shows which model a tool runs on.

May I use confidential data?

Never use AI as a source: every reference must be traceable to the original article, as fabricated or wrongly attributed citations are the most common error. Unpublished data, research proposals and peer-review manuscripts are confidential or commercially sensitive, so do not enter them into a consumer account without a data processing agreement, and check whether the provider stores your input or uses it for training (publishers usually prohibit this explicitly for peer review). If you work with personal data from research participants, the GDPR and your own ethics review apply: running models locally or using an EU-hosted tool with data minimisation is then the safe route. Finally, document your search strategy and your AI use: journals and funders increasingly ask for this, and a review must remain reproducible.

What does the GDPR profile mean?

Our summary of what the provider publishes about data protection: EU hosting, no training on your data, certifications and provenance. 'Strong' = several safeguards published; 'Unknown' = the provider shares little, so ask them. It is a tool, not legal advice.