We lay out the facts, we don't decide for you
Vroooz gathers public, sourced facts, then lays them out clearly. We don't tell you what to think: we give you what you need to form your own view. You're the one who decides.
Strict rule: we rate <strong>the claim</strong>, the message, never the named individuals. No ranking of people, no trial. A funding source is a fact we can cite; it isn't an accusation as long as we assert no cause-and-effect link.
We can be wrong, and science evolves. We'd rather say "we don't know yet" than rush to a call.
The 5 questions asked of every claim
Behind every check, we always answer the same questions. They're the ones you should really ask when facing a food claim.
How we find the evidence
When you submit a claim, we search the public scientific literature live, mainly <strong>Europe PMC</strong> and <strong>PubMed</strong>, the two big databases of biomedical papers.
We don't search at random. We reason with an <strong>"exposure AND effect"</strong> logic: we cross the food or substance in question with the claimed health effect. For "coffee causes cancer", say, we look for studies about both (coffee) AND (cancer), not just one or the other. That way, we only bring back studies that really speak to the alleged link.
We also look at what the study compares. "Diet soda versus nothing" and "diet soda versus sugary soda" aren't the same questions, so they don't give the same answers. If the study we find doesn't compare the right thing, we flag it instead of pretending otherwise.
We also rely on official guidelines
Some questions are already settled by public-health institutions. When a claim touches a validated benchmark, we rely on official guidelines, Belgium's Superior Health Council, the EU's EFSA and France's HAS, which have already synthesised meta-analyses. We then cite the source, why it exists, and which studies back it. And we always check it against the most recent studies.
And when no study tests a claim precisely but an established principle can answer it, we tell you and lean on it rather than dodging with an "it depends". For example, weight gain is multifactorial: it depends on your overall diet, physical activity, sleep, stress, genetics or medication, never on a single food or on calories alone.
We also apply the WHO's precautionary principle. When the International Agency for Research on Cancer (IARC) classifies a substance as carcinogenic, aspartame for instance, rated "possibly carcinogenic" (group 2B) in 2023, we flag it, distinguishing hazard from dose, even when an official acceptable daily intake exists and is debated.
Not all studies are equal
A study on a single person doesn't carry the same weight as an analysis combining dozens of studies. So we grade the strength of the evidence on four levels, from strongest to most fragile.
They combine several existing studies and rerun the numbers on the whole. Large pooled samples, a reliable overall trend.
A real intervention is tested, with participants split at random between groups. It's the best proof of a cause-and-effect link, but often on few people.
Large populations are observed over time. It shows an association, a risk, never a cause and effect.
A testimony, a case series, an editorial. Useful to open research leads, not to ground a recommendation.
<b>Important rule:</b> the level of evidence is capped by the weakest study we lean on to conclude. We can only show "High" if we hold at least one solid meta-analysis or several randomised trials pointing the same way. A single observational study never reaches "High".
Five verdicts, and one rule of caution
At the end of the analysis, we sum it up with a clear verdict, designed to be read at a glance.
<b>The rule that matters most:</b> when there aren't enough serious studies, we never say "false". We say "we don't yet have enough studies to decide". Absence of proof isn't proof of the opposite, it's a cautious position, not a denial.
Likewise, when a claim rests only on observation (an association, not a cause and effect), we avoid blunt "true" or "false" verdicts and add nuance.
Relative risk versus absolute risk
It's the classic headline trap. "Eating this raises the risk by 30%" says nothing concrete on its own. 30% more compared to what, out of how many people.
So whenever we can, we translate it into <strong>"out of 100 people"</strong>, which everyone understands without any grasp of statistics.
"Drinking one soda a day raises the risk of diabetes by 80%."
"Out of 100 people who don't drink it, about 5 will develop diabetes. Out of 100 who drink one a day, it's closer to 9, that is 4 more people per 100."
And when the study doesn't give the starting point (which happens often), we say so plainly rather than leaving an impressive percentage on its own.
The food matrix: the same nutrient, not the same effect
A sugar doesn't act the same whether it comes from a whole fruit or a juice. The whole food, the chewing, the fibre, the real dose, the speed of absorption: all of it changes the effect on your body.
An orange juice is the sugar of three or four oranges, without the fibre, downed in a few seconds. That isn't the same as biting into an orange. We surface this context nuance every time it matters.
What we rely on
The big public databases of scientific papers. That's where we go for the evidence, live.
To trace who funds what: lobbies, interest groups, official declarations.
The French health-safety agency, for official safety and public-health benchmarks.
The public-health guidelines of France's national nutrition and health programme, in plain language.
Belgium's official dietary recommendations (2025), graded by level of evidence.
The European food-safety authority: the EU's dietary reference values.
France's National Authority for Health, for validated public-health benchmarks (weight, physical activity).
The carcinogenic-agent classification of the WHO's International Agency for Research on Cancer, under the precautionary principle.
What we don't hide
We're building an honest tool, not an oracle. We can be wrong, a study can slip past us, and today's science will sometimes be corrected tomorrow. That's normal.
That's why we're building a way to <strong>report an error</strong> and a <strong>right of reply</strong>, so it gets fixed fast when it's warranted. Vroooz improves with you.