ARTICLE IMAGE
Nutrition
Could your morning coffee help you live longer?
A new study suggests a possible association between regular coffee consumption and...
Research for healthier information ecosystems
See the evidence
behind the headline.
HEALTH CLAIMS
NEWS MEDIA
NEWS ANALYSIS
NEWS CORPUS
CLAIM MINING
CLAIM EXTRACTION
TOPIC CATEGORISATION
MEDICAL NER
CONTENT ENRICHMENT
ADVERTORIAL DETECTION
CLAIM VERIFICATION
FACT CHECKING
EVIDENCE ALIGNMENT
TRUSTED SOURCES
EVIDENCE LAYER
SCIENTIFIC EVIDENCE
HEALTH LITERACY
RESPONSIBLE AI
TRUSTED JOURNALISM
PUBLIC HEALTH
INFORMED DECISIONS
MEDIA RESEARCH
MISINFORMATION
HEALTH INFORMATION
CONTEXT
TRANSPARENCY
TRUST
The challenge
Health information travels fast.
Context often does not.
Claims are numerous
Food, fitness, body image, treatments and public-health topics appear throughout news and social media.
Context gets lost
Uncertainty, study design, populations and important limitations may disappear as research becomes a headline.
Advertorial content
Health claims can be part of paid or commercially influenced content and require careful interpretation.
Real-world harm
Health misinformation can affect vaccination, diets, treatments, trust and collective public-health decisions.
Selected evidence
Numbers behind the argument.
of 18–24-year-olds say social media, video networks and AI are their main source of news.
Reuters Digital News Report 2026trust news most of the time globally; trust is 22% for news on social media and 20% for news from AI chatbots.
Reuters Digital News Report 2026of surveyed news leaders said generative AI was fully or somewhat transforming newsrooms.
Journalism, Media & Technology Trends 2025misinformation and disinformation ranked as the top global risk over the two-year horizon.
World Economic Forum Global Risks Report 2025The VITAL vision
Not another verdict engine. An explainable evidence layer.
VITAL is being developed as a Norwegian-built infrastructure for health and lifestyle news claims: a platform that analyses reporting, connects claims to evidence and context, and helps people understand uncertainty without reducing complex research to a simplistic true-or-false label.
Can an evening walk improve your sleep?
A fictional article suggests that an evening walk improves sleep quality and presents the finding as a general recommendation.
Illustrative concept only — not a real verification result.
Extract claims
Identify the health claims contained in news articles.
Grade evidence
Show the quality, scope and limitations of relevant evidence.
Offer alternatives
Connect readers to additional trusted sources and context.
Adapt communication
Present useful explanations for readers, journalists and researchers.
In practice
Designed for the moments
where context matters.
VITAL explores how the same research engine can support readers, journalists and researchers at different points in the information journey.
Reader experience
Understand a claim without leaving the story.
Readers can receive concise, clearly signposted context beside an article—helping them distinguish between what a study found and what a headline implies.
Notice uncertainty
See study type, limitations and scope.
Explore alternatives
Find additional context and relevant sources.
Make informed choices
Use evidence without requiring specialist expertise.
The VITAL approach
From health-news claims
to evidence and context.
News corpus
Build and maintain a large collection of health and lifestyle news articles.
INPUT
→
Claim mining & categorisation
Extract article topics and identify the health claims they contain.
EXTRACTION
→
Evidence alignment
Compare claims with trusted sources and relevant supporting evidence.
VERIFICATION
→
Communication
Present useful context, explanations and alternative sources to different users.
EVIDENCE LAYER
CORE RESEARCH QUESTION
Which health claims are present, how are they framed, what evidence do they align with, and how should this be communicated to different users?
Which claims are present?
Identify health claims across large collections of lifestyle and health news.
How are they framed?
Examine wording, sentiment, article categories and advertorial characteristics.
What evidence supports them?
Compare claims with trusted sources and assess whether evidence aligns with the reporting.
How should context be communicated?
Design an evidence layer that works for readers, journalists and researchers.
Hands-on example
See how an article
could become explorable.
The article, claims, evidence details and source references below are generated solely to demonstrate the intended interaction.
Two cups of coffee a day may protect your heart, new study suggests
Researchers following an example group of adults reported an association between coffee consumption and cardiovascular outcomes.
The researchers observed a difference between groups, although the study design cannot establish that coffee caused the outcome.
The article also reports that , which readers could interpret as a general recommendation.
An accompanying summary concludes that , despite possible differences between populations and individual health circumstances.
Current research
A growing research engine
for health-news analysis.
The current engine works with a processed subset of the wider collected dataset. Research outputs continue to evolve as datasets, methods and evaluations are refined.
processed news articles currently available in the research engine
health and lifestyle articles collected for research
accuracy reached by the fine-tuned ModernBERT model
Data collection
NewsCatcher API and corpus construction
Cleaning & enrichment
Media Bias Fact Check, NER, Medical NER, sentiment, embeddings and advertorial detection
Claim mining
Claim extraction, Factiverse API and article categorisation
Data analysis
Current English-language health-news study
Research outputs
Preliminary findings
Early signals from
the ongoing study.
These findings are preliminary and should not be interpreted as final project conclusions.
Advertorial content scores higher on factual accuracy
Current analysis suggests that advertorial health content scores higher on factual accuracy than regular health-news content.
Positive articles show more supported claims
Positive articles are associated with a higher share of supported health claims.
Accuracy varies by medical topic
Factual accuracy scores differ significantly between the medical topics covered by news articles.
The people behind VITAL
The current
VITAL team.
1 / 1
Collaboration direction
Research strengthened
through collaboration.
VITAL’s research engine, claim explanations and evidence-layer research can be combined with Factiverse’s AI credibility checks, claim detection and verification capabilities.
VITAL research engine
+
Factiverse verification
A Norwegian, research-based health-information resilience stack
Research environment
Verification technology
Related publications
Related research from
MediaFutures.
1 / 1
Stay connected