Evidence Intake
Run an analysis
Analysis is language-based, not a database fact-check — it flags vague vs. specific claims and explains its reasoning. It does not verify certifications independently.
Greenified reads sustainability claims — pasted text or a photo of a label — and shows exactly which parts are backed by specifics, and which are just green-sounding words.
Greenwashing is when marketing makes a product or company sound more sustainable than the substance behind it actually supports — think "eco-friendly" with no explanation, or "carbon neutral by 2030" with no real plan. It matters for SDG 12 (Responsible Consumption) and SDG 13 (Climate Action) because it lets companies claim credit for sustainability without doing the harder work.
Paste an ad claim or upload a photo of a label, product, or ad.
The AI checks what it's actually looking at, then reads each sustainability claim.
Every claim is marked specific, vague, or unverifiable — with plain-language reasoning, not just a score.
Built for OurPlanet.Rocks — a student hackathon tackling the UN Sustainable Development Goals.
Greenified runs a three-stage reasoning pipeline, not a single black-box prompt — every step is designed to be auditable rather than a trust-me output.
Before any analysis runs, the model first checks what it's actually looking at — a real product or ad with identifiable claims, an unrelated image, something too ambiguous to read, or a vague mention with nothing concrete to evaluate. Most AI tools skip this and hallucinate confidence where none is warranted. If the input names a well-known brand without pasting the exact claim, the model draws on general knowledge of that brand's documented practices instead of stalling — and says clearly when it's doing so.
Rather than one opaque score for a whole product, each individual claim is tagged on its own: specific (backed by a named certification, a concrete number, or a verifiable mechanism), vague (feel-good language like "eco-friendly" with nothing behind it), or unverifiable (plausible, but not assessable from what's given). Every tag comes with a one-sentence, plain-language reason.
The 60–100% score is anchored to fixed reference bands, not a gut feeling — 95–100% requires near-total specificity with no contradicting history, while 60–64% requires near-total vagueness or documented contradicting evidence like a past scandal or regulatory action. The model counts the actual ratio of specific-to-vague claims, matches it to the right band, then adjusts for known contradicting evidence — turning a subjective judgment into a structured, comparable one.
The same pipeline runs on pasted text or a photo of a physical label — vision and language analysis feed into the same evidence-tagging logic, so a shampoo bottle photo gets the same rigor as a typed-in ad claim.
Greenified isn't a live fact-checking database — it can't verify a certification is currently valid the way a regulator could. It's built to do what a careful, literate reader would do on a first pass: separate substantiated claims from marketing language, and say plainly when it doesn't know enough to judge.
Greenwashing isn't hypothetical — regulators and journalists have caught major companies doing exactly this, and it happens at every scale. The most famous cases involve billion-dollar brands, but the same vague-label tricks show up on a $12 bottle of "natural" shampoo at a corner store just as often as on a Fortune 500 sustainability report. A few well-documented big-name examples first:
For years, VW advertised its diesel engines as low-emission. In September 2015, U.S. regulators found the real mechanism: software that detected when a car was being emissions-tested and temporarily altered engine performance to pass — while emitting up to 40 times the legal nitrogen oxide limit on the road. Roughly 11 million vehicles worldwide were affected. This is the clearest case on this list: the claim wasn't just exaggerated, it was engineered to deceive a specific test.
In 2018, McDonald's UK swapped plastic straws for paper ones, framing it as a plastic-reduction win. The problem surfaced a year later: the paper was too thick for UK recycling machinery to process, meaning the "eco" straws were going to landfill just like the plastic ones — while the swap did nothing about McDonald's much larger sources of packaging waste. It's a textbook "hidden trade-off": a visible fix for a small piece of the problem, standing in for the harder work on the rest of it.
Coca-Cola Life launched with green-labeled bottles and ad imagery built entirely around nature and health, positioning it as the natural, better-for-you Coke. But the "green" was mostly the label: it still carried meaningful added sugar, and nutritionists said the branding overstated how different it really was from regular Coke. Consumer confusion over what the product actually was — a diet drink? a healthier original? — contributed to Coca-Cola pulling it from shelves in 2017, three years after launch.
Small brands do it too — usually more crudely, and with far less accountability, since they rarely face a regulator or a journalist checking their claims. A shampoo bottle that says "natural" with no ingredient breakdown. A candle brand calling itself "eco-conscious" with no certification, no sourcing info, nothing but the word itself. A local coffee bag stamped "sustainably sourced" with no indication of what standard that's measured against. These claims get less scrutiny than Volkswagen or Coca-Cola precisely because nobody's watching closely enough to catch them — which is exactly the gap this tool is meant to help close.
Most people don't have time to research every brand's claims the way investigative journalists or regulators do. Greenified gives anyone the same first pass — read a claim, and see whether it's backed by something real or just sounds good. The goal isn't to accuse companies; it's to make vague, unverifiable language easier to spot before it does its job.
So I built this app to raise people's awareness of greenwashing — and made it so anyone can check whether a company or product actually has real evidence behind its claims, instead of just taking the label's word for it.
This isn't just a hackathon talking point — regulators, UN officials, and climate advocates have been saying it publicly for years. A few statements on the record:
Analysis is language-based, not a database fact-check — it flags vague vs. specific claims and explains its reasoning. It does not verify certifications independently.