How we verify a claim, evidence first
We don't take a brand's word for it, and we're not asking you to take ours either. This page walks through exactly how TilliT checks a claim like Organic or Vegan. It's also honest about the parts we're still working on.
The process, step by step
Six steps, every time, for every product, whether or not the brand has ever heard of TilliT.
1.We work out exactly what's being claimed
“Organic” and “vegan” aren't single, simple ideas. What counts as proof depends on where a product is sold and how the claim is worded. So we start by pinning down exactly what would need to be true.
2.We gather evidence outside the brand's own control
Official certification registries, government accreditation records, credible news coverage. All of it is stuff a shopper, a journalist or a regulator could go and check themselves.
3.We weigh the evidence with a fixed, published formula
Not a gut feeling, and not an unexplained AI decision. The same evidence produces the same result, for every product, every time.
4.A trained person checks the outcome
Every case moves through human review before it's treated as final, and higher-stakes situations are automatically sent for extra, independent confirmation.
5.Everything is written down permanently
Every piece of evidence, every reviewer decision, and the reason behind it, gets recorded in a way that can't be quietly edited or deleted. You can only add to it.
6.We keep checking after the badge is issued
Certifications lapse, records change, and news shifts what's known about a company. So we keep rechecking, not just once at the start.
Who relies on this
We wrote this methodology to be useful well beyond shoppers checking a single product:
This list is deliberately broad, and it isn't the limit. If you have any stake in whether a product claim can be trusted, this page is for you too.
The principles behind every check
Seven habits we apply to every single assessment, no exceptions.
Where the evidence comes from
Only sources a third party could check themselves: official registries, government records, credible news. A brand's own site can back up a claim, but it can't establish one on its own.
Cross-checking evidence
A claim backed by more than one genuinely independent source counts for more than several mentions that all trace back to one origin.
Keeping our incentives honest
No part of scoring ever takes a merchant's identity or billing status into account. Reviewers declare conflicts of interest and are a separate team from account management.
Read our full independence policyFixing mistakes
Every scoring change has to pass 45 real, worked test cases before it can go live, and an automatic check flags rules reviewers keep overturning.
How sure we really are
Confidence levels come from a fixed formula, not a typed-in impression. And we track how often our own first read matches the final human call.
Staying current
Certifier status is rechecked on a set schedule. A clean result gets recorded as a real finding too, not thrown away just because nothing was wrong.
A secure, open system
Every internal endpoint is protected against misuse, and every result is stamped with the exact version of our method that produced it.
Human oversight, done rigorously
Blind second reviews, measured reviewer agreement, and mandatory sign-off on the highest-stakes corrections. It's modelled on how clinical and scientific research handles the same problem.
See how it follows the scientific methodWhat we check, badge by badge
Each badge is checked differently, because each claim is regulated and certified differently in the real world.
Organic
“Organic” isn't one legal claim. It shifts by market and by wording. We work out which claim applies, then check it against the labeling rules of six major jurisdictions (US, EU, Canada, Japan, South Korea, Switzerland), confirm the certifying body's accreditation is currently in good standing, and check a historical record of certifiers that have run into accreditation problems before.
See organic products we've checkedVegan
Screened against a maintained list of animal-derived ingredients and their many lesser-known names, in French, German and Spanish as well as English. We test it against real product listings so a plant-based ingredient isn't mistaken for an animal-derived one that sounds similar.
See vegan products we've checkedNon-GMO
Cross-checked against the leading independent non-GMO verification program's own records, plus a maintained list of crops commonly genetically modified. A claim paired with a high-risk ingredient gets flagged for closer review.
See non-gmo products we've checkedGluten-Free
Checked directly against recognized, independent gluten-free certification bodies, rather than inferred from an ingredient list. A certificate is confirmed against the certifier's own records first.
See gluten-free products we've checkedAllergen-Free
Checked against a live, public allergen-free certification listing. We hold ourselves to a strong-but-not-absolute confidence standard here, and we say exactly how sure the underlying public data lets us be.
See allergen-free products we've checkedFair Trade
Checked against a live registry tracking tens of thousands of registered fair-trade products and producers, alongside the certifying organization's own standing. Public data doesn't yet support a guaranteed exact match to a single barcode, and we say so rather than round up.
See fair trade products we've checkedWhat we're still working to improve
No verification process is ever finished, and a trustworthy one should say so. These aren't things we've overlooked. They're areas where doing it honestly takes longer than the rest of the system did, given where TilliT is right now.
Empirical validation against real-world ground truth
Our scoring weights come from careful, documented design work. But they haven't been formally validated yet against a large, independently labelled dataset of real products with known compliance status. That's a hard, data-heavy job. We're building toward the benchmark dataset it needs rather than pretending the current design is a finished answer. Doing it properly means gathering a lot of products whose true status is independently and reliably known, and that takes time. Rush it and you undermine the validation it's supposed to provide.
Language and regional coverage
Our global news and enforcement monitoring already covers multiple regions and nine languages. Detecting the product claim itself is most mature for English, and badge by badge it's furthest along for vegan. We treat each new language as its own tested build, not a quick translation. We'd rather add one language at a time and get it right than claim broad coverage that isn't equally reliable everywhere.
Visual logo verification
Matching a certification logo in a photo is one supporting signal among several, and it's the kind of thing that gets better with more real-world examples. While we build that up, we weight it conservatively. A document or registry match always counts for more than a logo match on its own. That's on purpose, so we don't lean too hard on a signal that's still maturing.
How to read and use a TilliT report
What a badge actually represents, and just as importantly, what it doesn't.
What a badge means: a specific claim was checked against public evidence, as of a certain date. It's not a lifetime guarantee, and it doesn't take over the brand's own legal responsibilities.
Look at the date: evidence changes. A badge tied to a recent check carries more weight than one that hasn't been revisited in a while.
“We couldn't confirm this” is a real answer: not a red flag by itself. Some claims just don't have enough public evidence yet, especially for smaller or newer brands.
Treat it as one strong input, not the only one: evidence-based and independently checkable, but still one input alongside your own research and judgment.
If something looks wrong, tell us: every result comes with a way to raise a concern, and we track corrections permanently.
TilliT verifies, not certifies: we check whether a claim is backed by evidence from certifying bodies and public records. We aren't a certifying body ourselves.
A badge from TilliT means a specific, checkable claim was tested against public evidence using a fixed, published method. Not a brand's word, and not some opaque AI's unexplained opinion.