From +10,000 articles to your shortlist — in hours, not months
Independent AI reviewers evaluate every record against every criterion — so you can see where models agree, where they disagree, and why.
Sign up for free and screen up to 500 records with our trial.
Four steps to your evidence shortlist
No coding. No setup. No machine learning expertise required.
Upload your dataset
Import your references from PubMed, Scopus, Web of Science, or any database. We support RIS, BibTeX, NBIB, and CSV formats.
Define your criteria
Set exclusion and characterization criteria in plain language. Our AI assistant helps you refine them using PICO frameworks.
AI panel screens every record
Models evaluate independently. If they disagree, they exchange anonymized arguments across multiple rounds until consensus is reached — or flagged for your review.
Review the consensus
Get your shortlist ranked by agreement rate. See which models agreed, which disagreed and why — for every single record.
Your criteria can evolve.
Every record is evaluated against every criterion — nothing is dropped early. If you change your mind about which criteria should exclude articles, you get new results in seconds. No reprocessing.
What used to take months of reading now takes hours. You get the articles that matter — we handle the rest.
One AI model decides. Multiple models deliberate.
Inspired by the Delphi method: independent judgments, anonymous feedback, and structured re-evaluation when reviewers disagree.
Traditional AI screening
One algorithm makes all decisions
Round 1: models evaluate independently
No way to measure consensus
Agreement rate quantifies inter-model consensus
Errors invisible until human review
Disagreement triggers a new round with anonymized peer arguments.
Black-box recommendations
Full justification per record, per criterion, per reviewer, per consensus
Change one criterion, reprocess everything
Change which criteria exclude — new results in seconds, same evaluations.
“When 4 out of 5 AI reviewers agree a paper should be excluded — and each gives a different reason aligned with your criteria — you can trust the decision. When they disagree, you know exactly where to focus your review.
Frontier AI Models. Working Together.
Choose from 40 AI models across 8 providers. Different architectures and error patterns make disagreement your most valuable signal.
You choose the panel. Independent reviewers make uncertainty visible.
Built on proven research methodology
Every claim on this page is backed by peer-reviewed research — not marketing promises.
The Delphi Method
Used in over 10,000 published studies to reach reliable consensus among independent experts. We brought large language models into the expert panel — same rigor, faster results, with you always in control.
The Cost of Manual Screening
Manual dual-screening of 10,000 records takes ~667 researcher-hours. AI-assisted tools reduce that by 50–80%.
Transparent by Design
The scientific community demands reproducibility. Every Delph-AI decision includes the full reasoning chain — which models agreed, which disagreed and why.
“A methodology validated for 75 years — and the first implementation that brings it to AI-powered screening.
The first systematic review tool in your language
Our AI models understand and screen literature in 90+ languages — your dataset, in any language.
English
Screen thousands of papers with AI consensus
Español
Filtra miles de artículos con consenso de IA
Português
Triagem de milhares de artigos com consenso de IA
Français
Filtrez des milliers d'articles avec le consensus IA
Deutsch
Tausende Artikel mit KI-Konsens screenen
中文
使用AI共识筛选数千篇文献
日本語
AIコンセンサスで数千の論文をスクリーニング
العربية
فحص آلاف الأوراق بتوافق الذكاء الاصطناعي
한국어
AI 합의로 수천 편의 논문 스크리닝
Italiano
Screening di migliaia di articoli con consenso IA
Nederlands
Screen duizenden artikelen met AI-consensus
Русский
Скрининг тысяч статей с помощью ИИ-консенсуса
English
Screen thousands of papers with AI consensus
Español
Filtra miles de artículos con consenso de IA
Português
Triagem de milhares de artigos com consenso de IA
Français
Filtrez des milliers d'articles avec le consensus IA
Deutsch
Tausende Artikel mit KI-Konsens screenen
中文
使用AI共识筛选数千篇文献
日本語
AIコンセンサスで数千の論文をスクリーニング
العربية
فحص آلاف الأوراق بتوافق الذكاء الاصطناعي
한국어
AI 합의로 수천 편의 논문 스크리닝
Italiano
Screening di migliaia di articoli con consenso IA
Nederlands
Screen duizenden artikelen met AI-consensus
Русский
Скрининг тысяч статей с помощью ИИ-консенсуса
English
Screen thousands of papers with AI consensus
Español
Filtra miles de artículos con consenso de IA
Português
Triagem de milhares de artigos com consenso de IA
Français
Filtrez des milliers d'articles avec le consensus IA
Deutsch
Tausende Artikel mit KI-Konsens screenen
中文
使用AI共识筛选数千篇文献
日本語
AIコンセンサスで数千の論文をスクリーニング
العربية
فحص آلاف الأوراق بتوافق الذكاء الاصطناعي
한국어
AI 합의로 수천 편의 논문 스크리닝
Italiano
Screening di migliaia di articoli con consenso IA
Nederlands
Screen duizenden artikelen met AI-consensus
Русский
Скрининг тысяч статей с помощью ИИ-консенсуса
English
Screen thousands of papers with AI consensus
Español
Filtra miles de artículos con consenso de IA
Português
Triagem de milhares de artigos com consenso de IA
Français
Filtrez des milliers d'articles avec le consensus IA
Deutsch
Tausende Artikel mit KI-Konsens screenen
中文
使用AI共识筛选数千篇文献
日本語
AIコンセンサスで数千の論文をスクリーニング
العربية
فحص آلاف الأوراق بتوافق الذكاء الاصطناعي
한국어
AI 합의로 수천 편의 논문 스크리닝
Italiano
Screening di migliaia di articoli con consenso IA
Nederlands
Screen duizenden artikelen met AI-consensus
Русский
Скрининг тысяч статей с помощью ИИ-консенсуса
English
Screen thousands of papers with AI consensus
Español
Filtra miles de artículos con consenso de IA
Português
Triagem de milhares de artigos com consenso de IA
Français
Filtrez des milliers d'articles avec le consensus IA
Deutsch
Tausende Artikel mit KI-Konsens screenen
中文
使用AI共识筛选数千篇文献
日本語
AIコンセンサスで数千の論文をスクリーニング
العربية
فحص آلاف الأوراق بتوافق الذكاء الاصطناعي
한국어
AI 합의로 수천 편의 논문 스크리닝
Italiano
Screening di migliaia di articoli con consenso IA
Nederlands
Screen duizenden artikelen met AI-consensus
Русский
Скрининг тысяч статей с помощью ИИ-консенсуса
English
Screen thousands of papers with AI consensus
Español
Filtra miles de artículos con consenso de IA
Português
Triagem de milhares de artigos com consenso de IA
Français
Filtrez des milliers d'articles avec le consensus IA
Deutsch
Tausende Artikel mit KI-Konsens screenen
中文
使用AI共识筛选数千篇文献
日本語
AIコンセンサスで数千の論文をスクリーニング
العربية
فحص آلاف الأوراق بتوافق الذكاء الاصطناعي
한국어
AI 합의로 수천 편의 논문 스크리닝
Italiano
Screening di migliaia di articoli con consenso IA
Nederlands
Screen duizenden artikelen met AI-consensus
Русский
Скрининг тысяч статей с помощью ИИ-консенсуса
English
Screen thousands of papers with AI consensus
Español
Filtra miles de artículos con consenso de IA
Português
Triagem de milhares de artigos com consenso de IA
Français
Filtrez des milliers d'articles avec le consensus IA
Deutsch
Tausende Artikel mit KI-Konsens screenen
中文
使用AI共识筛选数千篇文献
日本語
AIコンセンサスで数千の論文をスクリーニング
العربية
فحص آلاف الأوراق بتوافق الذكاء الاصطناعي
한국어
AI 합의로 수천 편의 논문 스크리닝
Italiano
Screening di migliaia di articoli con consenso IA
Nederlands
Screen duizenden artikelen met AI-consensus
Русский
Скрининг тысяч статей с помощью ИИ-консенсуса
English
Screen thousands of papers with AI consensus
Español
Filtra miles de artículos con consenso de IA
Português
Triagem de milhares de artigos com consenso de IA
Français
Filtrez des milliers d'articles avec le consensus IA
Deutsch
Tausende Artikel mit KI-Konsens screenen
中文
使用AI共识筛选数千篇文献
日本語
AIコンセンサスで数千の論文をスクリーニング
العربية
فحص آلاف الأوراق بتوافق الذكاء الاصطناعي
한국어
AI 합의로 수천 편의 논문 스크리닝
Italiano
Screening di migliaia di articoli con consenso IA
Nederlands
Screen duizenden artikelen met AI-consensus
Русский
Скрининг тысяч статей с помощью ИИ-консенсуса
We know which models we use. And so do you.
Transparency isn't a feature — it's our architecture.
AI Transparency
See which models reviewed each record, how they voted, and why. No black boxes. Built with EU AI Act Art. 50 transparency principles.
Data Protection
Encrypted in transit and at rest. Core application infrastructure hosted in EU (Belgium). AI inference via multiple providers (see Sub-processors). Built with GDPR principles in mind. Delph-AI never uses your data to train AI models.
PRISMA-ready audit trail
Export every model vote, criterion-level rationale per deliberation round, and final decision. Designed to support transparent, reproducible PRISMA 2020 reporting.
Compliance
GDPR principles, PCI-DSS (via Dodo Payments), LFPDPPP (Mexico), LGPD (Brazil). We publish our security practices openly.
Choose a plan that fits your research. Scale up when you need to.
Transparent pricing. No hidden fees.
Need more records? Purchase additional record packs anytime — starting at $1239.
Screen faster. Keep every decision traceable.
Choose your AI reviewer panel, surface disagreement, and focus human review where it matters most.