The problem
In 2020 and 2021, thousands of scientific publications emerged in just a few months on a single topic. Some studies claimed a drug worked. Others claimed the opposite. All were "published". All had "numbers". All could be cited.
Faced with this avalanche, journalists, doctors, and patients had no tools to cut through the noise. The question was no longer "is there a study that says that?" — there always was. The real question was: is this study worth anything?
Scientific misinformation is not always a problem of bad faith. It is often a problem of missing tools. When nobody has the means to evaluate the quality of evidence, any publication can pass for a truth.
Our answer: measure the process, not the results
Yuka does not evaluate whether you will like the taste of a yogurt. It evaluates its composition: sugar, additives, salt. The Nutri-Score does not say whether a recipe is good — it says whether it is balanced according to documented nutritional criteria.
Publi-Score does the same for scientific articles. We evaluate the process that produced the results:
- ·Was the study pre-registered before it started, or were the hypotheses formulated after seeing the data?
- ·Are the raw data accessible for independent verification?
- ·Is the funding independent of the manufacturer of the tested product?
- ·Is the comparison group appropriate? Is the sample large enough?
We do not say whether a drug works. We do not say whether the authors are right. We say whether the study followed the rules of the game that make a result credible.
A thermometer can be accurate or inaccurate — that has nothing to do with the temperature it measures. Similarly, a study can have true results and a fragile methodology, or false results and an impeccable methodology. Publi-Score measures the quality of the thermometer, not the temperature.
Three non-negotiable principles
Free forever
The quality of scientific evidence cannot be behind a paywall. Knowing whether a study is methodologically sound is civic information, not a premium service. The quick score will always be free, no account required, no limits.
Independent
No advertising. No industrial funding. No institution deciding the score. The scoring grid is public, documented, and contestable. Our funding comes solely from Premium subscriptions — not from the actors whose research we evaluate.
Directionally neutral
The conclusion of a study — positive, negative, significant or not — never influences its Publi-Score. A well-conducted study showing that a drug does not work gets the same score as a well-conducted study showing that it does. It is the process that counts, not the result.
What the grid is built on
The Publi-Score v1 grid is built on internationally recognized standards: CONSORT criteria for randomized controlled trials, PRISMA for meta-analyses and systematic reviews, the RoB 2 tool (Risk of Bias) from the Cochrane Collaboration, and GRADE recommendations for evaluating the level of evidence.
The grid covers 7 categories and 30+ sub-criteria. It was validated on a corpus of well-known publications (reference trials, Cochrane meta-analyses, controversial studies) and registered with e-Soleau to date its creation.
It is not perfect. No grid is. Our limitations are documented publicly on the transparency page. We acknowledge them and correct them iteratively.
What comes next
Publi-Score has been publicly available since March 2026. The next steps are, in order: enriching the catalogue with hundreds of AI-scored publications, opening institutional access (labs, newsrooms, health agencies), and progressively extending the grid to disciplines beyond biomedicine.
If you believe that science deserves better than simplified headlines, we think the same thing. Try the tool on an article that matters to you.