Green Claims Research: How to Substantiate a Sustainability Claim With Consumer Perception Evidence

Answer first: Every environmental marketing claim carries two separate substantiation burdens, and they need completely different evidence. The science burden asks whether the underlying environmental fact is true, and is answered with testing, lifecycle assessment, or carbon accounting. The perception burden asks what the claim actually conveys to a reasonable consumer, including implied meanings you never intended, and can only be answered by consumer research. Companies invest heavily in the first and almost nothing in the second, which is precisely backwards: regulators overwhelmingly bring greenwashing cases about what a claim implied, not about whether a number was wrong.

This guide explains the perception burden, shows what regulators say the test is in their own words, and gives you a research protocol that produces evidence capable of discharging it.

The two-burden model

Consider a claim that is entirely, verifiably true: our packaging is 15 percent lighter than our previous packaging. The science burden is trivially satisfied. You have the weights.

Now consider what a shopper takes away. The FTC Green Guides address almost exactly this fact pattern. A marketer states its packaging is Greener than our previous packaging, having reduced the weight by 15 percent with no other improvement. The Guides call that claim deceptive, because reasonable consumers likely would interpret Greener in this context to mean that other significant environmental aspects of the packaging also are improved. The number was true. The claim was still deceptive.

That gap between a true fact and a misleading takeaway is the perception burden, and it is a purely empirical question about what real people infer. You cannot reason your way to the answer from inside the marketing team, because the whole failure mode is that the people who wrote the claim know what they meant.

What the regulators actually say

United States. The FTC Green Guides sit at 16 CFR Part 260. Section 260.2 is the operative passage and it is worth reading closely. To determine if an advertisement is deceptive, marketers must identify all express and implied claims that the advertisement reasonably conveys, and must ensure that all reasonable interpretations of their claims are truthful, not misleading, and supported by a reasonable basis before they make the claims. In the context of environmental marketing claims, a reasonable basis often requires competent and reliable scientific evidence, defined as tests, analyses, research, or studies that have been conducted and evaluated in an objective manner by qualified persons and are generally accepted in the profession to yield accurate and reliable results.

Two phrases in that passage create research obligations. All implied claims that the advertisement reasonably conveys means you have to enumerate inferences you did not write. Before they make the claims means the evidence has to exist prior to dissemination; assembling it after a challenge is not substantiation.

Section 260.1(d) adds the two tests that shape study design. Whether a particular claim is deceptive depends on the net impression of the advertisement, label, or other promotional material at issue. And the Guides are based on marketing to a general audience, but where a marketer targets a particular segment of consumers, the Commission will examine how reasonable members of that group interpret the advertisement. The first sentence tells you to test the finished asset rather than the claim in isolation. The second tells you your sampling frame is a legal variable, not a convenience decision.

Section 260.4 deals with general environmental benefit claims and is unusually direct: because it is highly unlikely that marketers can substantiate all reasonable interpretations of these claims, marketers should not make unqualified general environmental benefit claims. The Guides give the brand name Eco-friendly as a worked example of a deceptive claim, because it likely conveys far-reaching environmental benefits and may convey that the product has no negative environmental impact.

Section 260.3(c) prohibits overstatement, with a memorable example: a rug labelled 50 percent more recycled content than before, where the content went from 2 percent to 3 percent. Technically true, and the Guides say it likely conveys the false impression that the manufacturer has increased significantly the use of recycled fiber.

Section 260.6(c) closes the certification loophole: third-party certification does not eliminate a marketer's obligation to ensure that it has substantiation for all claims reasonably communicated by the certification.

United Kingdom. The Digital Markets, Competition and Consumers Act 2024, whose unfair commercial practices provisions came into force on 6 April 2025, restates the net impression test in stronger language. Section 226(1)(b) makes it a misleading action to use an overall presentation which is likely to deceive the average consumer, and section 226(3) says flatly that an overall presentation may be deceiving even if the information it contains is true. That single subsection removes the but it was technically accurate defence from UK law.

Schedule 20 of the same Act lists practices that are unfair in all circumstances, with no need to prove any consumer was misled. Four are directly relevant to environmental claims: claiming to be a signatory to a code of conduct when the trader is not; claiming that a code of conduct has an endorsement from a public or private body which it does not have; displaying a trust mark, quality mark or equivalent without having obtained the necessary authorisation; and claiming that a trader, practice or product has been approved, endorsed or authorised by a public or private body when the claim is false or the terms of the approval are not being complied with. Eco-label misuse is not a grey area in UK law. It is a per se breach.

How common the failure is

In January 2021 the European Commission and national consumer authorities published the results of the first annual EU-level sweep focused on environmental claims, covering garments, cosmetics and household equipment. In 42 percent of cases the claims were found to be exaggerated, false or deceptive and could potentially qualify as unfair commercial practices. On the same day, a parallel global website sweep led by the UK Competition and Markets Authority and the Netherlands Authority for Consumers and Markets, coordinated through ICPEN, reported that 40 percent of firms green claims online could be misleading consumers, citing vague terms such as eco, sustainable and natural used without adequate explanation or evidence.

Those two numbers are the reason the perception burden matters commercially. If roughly four in ten claims in a sweep are challengeable, the base rate of a problem in your own portfolio is not small, and the cheapest time to find it is before publication.

Designing a claim perception study

The protocol below is the same in structure whether you run it as a survey or as a set of AI-moderated conversational interviews. What matters is the order.

Step 1: Show the finished asset, not the claim. The legal test is net impression, so the stimulus must be what the consumer will actually see: the pack, the product page, the ad, with imagery, colour, and any seals in place. Testing a claim string in a text box measures something the regulator will not be judging.

Step 2: Ask unaided first, always. This is the single design decision that determines whether the study is worth anything. Before any option list, ask an open-ended question such as: In your own words, what is this product telling you about its environmental impact? followed by What, specifically, do you think the company is promising here? Unaided elicitation is the only way to discover inferences you did not anticipate, and those unanticipated inferences are exactly the implied claims section 260.2 makes you responsible for.

The moment you present a checklist of possible meanings, you stop measuring inference and start measuring recognition. A participant shown the option this product is made from recycled materials will endorse it at a rate that has nothing to do with whether they would have thought of it. Aided measurement has a role, but only after the unaided phase has generated the option set.

Step 3: Probe the strength and specificity of each inference. For each meaning a participant volunteers, find out how firmly they hold it and how specific they think it is. Does 15 percent lighter mean only lighter, or also recyclable? Does the leaf icon refer to the product, the packaging, or the company? Section 260.3(b) requires claims to specify whether they refer to the product, the packaging, or a service, so this is a compliance question and not just a curiosity.

Step 4: Run a no-claim control cell. Half your sample sees the asset with the environmental claim, half sees it without. The difference in what people believe about the product is what the claim actually added. Without a control you cannot separate the claim's effect from the beliefs people already hold about the category, and category priors for things like bamboo, plant-based or refillable are strong.

Step 5: Test materiality. Deception requires the representation to be material to consumers decisions. Ask whether the claim changes what they would choose, and what they would expect to pay. A claim that shifts nothing is a lower risk claim.

Step 6: Sample the audience you target. If the campaign targets a specific segment, section 260.1(d) says the Commission will examine how reasonable members of that group interpret it. Your sampling frame therefore has to match your targeting, and that has to be documented.

Step 7: Test the qualification, not just the claim. If your compliance position rests on a qualifier, the qualifier is part of the stimulus and needs its own evidence. Section 260.3(a) requires qualifications and disclosures to be clear, prominent and understandable, placed in close proximity to the qualified claim, without distracting elements that could undercut it. Whether your footnote achieved that is measurable: show the asset, then ask unaided what the environmental claim was limited to.

Mapping the protocol onto structured questions

Koji supports six structured question types, and a perception study uses almost all of them in a deliberate sequence:

The open_ended step is the one that cannot be skipped or reordered, and it is the step that is hardest to run well at scale with a traditional survey tool. A text box collects a sentence and stops. What you actually need is a follow-up: when a participant writes better for the environment, somebody has to ask better than what, and in what way? That probe is where the implied claim surfaces.

This is what an AI-moderated interview does that a form cannot. Koji's AI interviewer asks the unaided question, reads the answer, and generates its own follow-up probes in the same conversation, in voice or in text, across every participant simultaneously. You get the depth of a moderated cognitive interview at the sample size of a survey, which is the exact combination a perception study needs and the exact combination that was previously unaffordable. Platforms built for forms, such as SurveyMonkey, Typeform and Qualtrics, capture the first answer and never ask the second question, which is where the evidence lives.

Because every interview is transcribed and analysed automatically, the output is also the documentation. You get the verbatim range of inferences, the frequency of each, and quotes tied back to the participant and the stimulus version they saw.

Documenting perception evidence

The point of the study is to produce a file you can hand to counsel or a regulator. It should contain the exact stimulus tested with version control, the sampling frame and why it matches the targeted audience, the unaided protocol showing that open questions preceded aided ones, the full range of inferences elicited with frequencies, the control-cell comparison, the materiality result, the date the research completed relative to the date of first publication, and the decision made as a result. That last item matters more than teams expect. Evidence that you ran the study and then changed the claim is far stronger than evidence that you ran it and shipped anyway.

Build this into your pre-launch review rather than treating it as a legal fire drill, in the same way our guide to research peer review describes catching design problems before a study fields.

A short list of claims that need perception testing before publication

• Any unqualified general benefit term: green, eco, sustainable, environmentally friendly, conscious, responsible
• Any comparative without an explicit basis: greener, cleaner, less waste, more recycled content
• Any seal, leaf, badge or colour system that is not a recognised certification with published criteria
• Any free-of or non-toxic claim, which the Guides treat as conveying benefits for both humans and the environment
• Any carbon, offset or net-zero claim, where consumer understanding of the mechanism is typically far weaker than marketers assume
• Any recyclable or compostable claim where real-world facility availability differs from technical capability

Frequently asked questions

What evidence do regulators expect behind a green claim?

Two kinds. Scientific evidence that the underlying environmental fact is true, which the FTC Green Guides describe as competent and reliable scientific evidence, meaning tests, analyses, research or studies conducted and evaluated in an objective manner by qualified persons. And evidence about what the claim conveys, because a claim is deceptive if it is likely to mislead reasonable consumers regardless of whether the literal words are true. The second kind is consumer perception research, and it is the one most companies never commission.

Why can I not just ask consumers whether a claim is misleading?

Because that measures opinion about misleadingness, not the inference the claim actually produces. The legal test is what a reasonable consumer takes away from the advertisement as a whole. To measure a takeaway you have to elicit it unaided, before showing any list of possible meanings. The moment you offer a checklist you are testing recognition of your own hypotheses rather than discovering what people independently inferred, and the unanticipated inferences are exactly the ones that create liability.

What is the net impression test?

Deception is judged on the overall impression an advertisement creates, not on any sentence in isolation. The FTC Green Guides state that whether a claim is deceptive depends on the net impression of the advertisement, label or other promotional material at issue. UK law puts it more bluntly: under section 226(3) of the Digital Markets, Competition and Consumers Act 2024, an overall presentation may be deceiving even if the information it contains is true. Practically, this means you test the finished asset with its imagery and seals, not the claim string alone.

Are unqualified claims like eco-friendly ever safe?

The Green Guides advise that marketers should not make unqualified general environmental benefit claims at all, because such claims likely convey far-reaching benefits and may convey that the product has no negative environmental impact, and it is highly unlikely a marketer can substantiate every reasonable interpretation. The Guides use the brand name Eco-friendly as a worked example of a deceptive claim. Qualification must be clear, prominent, in close proximity, and tied to a specific substantiated benefit.

How many participants do I need for a claim perception study?

It depends what you are measuring. Discovering the range of inferences a claim produces is a qualitative saturation problem and usually stabilises within 30 to 60 unaided conversational interviews per claim variant. Estimating the proportion of consumers who take away a particular meaning is a precision problem: at 95 percent confidence a sample of 400 gives roughly plus or minus 4.9 points and 1,000 gives roughly plus or minus 3.1. Run the qualitative phase first, because you cannot quantify a takeaway you have not yet discovered.

Does third-party certification remove the need for perception research?

No. The Green Guides state that third-party certification does not eliminate a marketer's obligation to ensure it has substantiation for all claims reasonably communicated by the certification. A seal can itself convey unintended meaning, such as implying the certifier tested the product when it only confirmed membership of an association. In the UK, displaying a trust mark or quality mark without the necessary authorisation, and claiming an endorsement a code of conduct does not have, are listed in Schedule 20 as unfair in all circumstances.

Related Resources

• Advertising Claim Substantiation covers the general framework for designing research that backs a marketing claim.
• Using Research Quotes in Marketing explains the FTC Endorsement Guides rules that apply once you put a participant quote in an ad.
• Structured Questions in AI Interviews details all six question types and when to use each in a perception protocol.
• Research Peer Review shows how to catch a leading or aided-first design before the study fields.
• Survey Question Types is a reference for question formats and their failure modes.
• Cross-Tabulation Analysis covers reading segment differences once you have quantified takeaway.

---

Need perception evidence before your next sustainability campaign ships? Start free with 10 credits and run an unaided claim study this week.