Vera, a case study from the ZOE Women in Tech Hackathon
Score the claim, not the influencer.
A mobile app for the beauty and wellness world that reads a supplement post the way you actually meet it, as a link, a screenshot, or a brand name, identifies the real product behind it, and checks what the evidence says. Built in two days.
The team
Who built Vera.
-
Irina CsapoLead Product Design & Engineering -
Patricia AmorimData Analyst, Business Insights & Data Visualisation -
Natasha IraniResearch Scientist, Data Science & Genetic Toxicology -
Rachel SillitoeMSc Human-Computer Interaction @ UCL, BSc Neuroscience -
Annabelle BodyHarvard MBA, ACA -
Meet AhluwaliaInnovation Strategist, Research, Technology & Culture, MA Innovation Management @ Central Saint Martins
The problem
Who do you trust with your body?
The global wellness industry is worth over five trillion pounds, and much of it runs on pseudoscience, cherry-picked studies, and influencer authority. Rigorous science does not perform well on Instagram. A charismatic creator with a supplement line does. Vitamins and supplements are now TikTok's single largest health and beauty category.
For someone scrolling, it is almost impossible to tell whether a claim is evidence-based, whether a headline has stretched a study, whether a product is backed by real data, or whether the advice is quietly shaped by a commercial incentive. When the loudest voice wins, people waste money, delay effective care, or make decisions about their bodies out of fear rather than evidence.
How might we help women under 35 evaluate the supplements their favourite creators are pushing, quickly, before they buy on vibe alone?
Why now
The moment is loud, and unregulated.
People are more health-conscious than ever, and the "one pill to fix it all" narrative has exploded across social media. Gen Z increasingly treat social platforms as their search engine, which makes them the most exposed to targeted misinformation, while the evidence that could actually help is locked behind paywalls or written in language no one reads.
The solution
Meet Vera.
Vera reads like a person's name, warm and easy to say, and it is rooted in the word "true." She is a friend, not an auditor. The whole app rests on one disarming principle: score the claim against the evidence, not the product or the person.
You paste in what you saw. Vera identifies the real product behind the post, breaks its promises down claim by claim, grades each one against the research, and links you to the studies. It is built for someone like Rachel, 27, who keeps getting served creatine-gummy ads and cannot tell what is worth her money.
We score the claim against the evidence, not the product or the person.
What it does
Three ways in, one honest answer.
However a claim reaches you, Vera meets it there. All three inputs flow into the same identification engine and land on the same rich results page.
Real product identity
The actual company, ingredients, dominant active ingredient, and a plain one-line summary, with source links.
Claim Score
An at-a-glance early estimate of how well the product's promises hold up, clearly labelled as a prototype.
Claim by claim
Each promise separated out and graded: strong, moderate, mixed, limited, or not supported.
Live study count
A real-time count of related peer-reviewed studies from OpenAlex, linked straight to the research.
Honest reviews
A sentiment-weighted review carousel that resists the miracle-story framing of the original post.
Better alternatives
Higher-scored products in the same category, so a "no" always comes with a "try this instead."
Under the hood
One engine, three ways in.
The interesting part is what happens after you paste. A link, a screenshot, and a typed brand name all resolve to the same question: what is the real product here, and what does the evidence say about it?
Reading the post
TikTok blocks the browser, so a small Cloudflare Worker fetches the public post server-side: the handle, follower count, date, a paid-partnership flag, and the cover image. It deliberately leaves out likes, which read as an endorsement the product has not earned.
Identifying the product
The caption and cover image, or your uploaded screenshot, or a typed brand name, go to Gemini with Google Search grounding. It returns the real company, the ingredient list, the dominant active ingredient, a factual one-line summary, and links to its sources. If it cannot verify something, that field comes back blank rather than invented.
Weighing the evidence
For the product's main ingredient, Vera queries OpenAlex live for the number of related peer-reviewed studies and links straight through to them, so the research is one tap away instead of behind a paywall.
The honest part
What is real, and what is still a prototype.
This matters more than any feature. We were hand-scoring results from our own literature review for the demo, and we said so out loud. Being precise about the line is what makes the rest credible.
- Live product identification from a link, image, or name.
- Real company, ingredients, and main ingredient, pulled from the web, never fabricated.
- OpenAlex study counts and source links, fetched in real time.
- The Claim Score is a heuristic, hand-built from our own reading of the evidence.
- A shipped product would need a defined, published scoring methodology.
- That methodology has to stay independent, including for ZOE's own products.
Where it goes next
From prototype to platform.
Questions we answered
The tough ones, from the judges.
How does Vera score a claim?
Is there legal risk in rating real brands?
How is this different from Yuka or Examine?
Why might ZOE score well, and is that a conflict?
What problem are we really solving?
What we took away
Six strangers, one working app, two days.
Team 11 came together at the ZOE Women in Tech Hackathon: two engineers, three product and strategy minds, and one designer, most of us meeting for the first time. We went from a blank problem statement to a working prototype that pulls real product data and real research, and, just as importantly, learned to be honest about where the prototype ends. It was a reminder that there is no one "right" path into tech, and that the room gets better when more of us are in it.
