How green are green claims?

Overview

Our new research sets out how businesses are making vague and unsubstantiated environmental claims. We use artificial intelligence to to identify green claims from product descriptions and to assess the claims made about a sample of 1,000 products against the Competition Market Authority's Green Claims Code.

You can read the full report below or download the PDF version. 

Executive summary

Green claims made by businesses can influence consumers’ decisions to purchase a product. However, consumers often struggle to verify the accuracy or authenticity of environmental claims, which creates an incentive for businesses to make misleading green claims. This can lead to consumers buying products they believe are more sustainable than they actually are, and paying a premium for them. It also undermines businesses that genuinely invest in sustainability, weakening their competitive advantage and discouraging other businesses from acting more sustainably. 

Despite there being recent regulatory action by the Competition and Markets Authority (CMA) and the Advertising Standards Authority (ASA), which suggests that misleading green claims may be common, there is little research on the prevalence of potentially misleading green claims in the UK. To address this, Which? and London Economics have undertaken research on the prevalence of green claims in the UK which may not comply with the CMA’s Green Claims Code, a guidance for businesses to ensure their green claims are clear, accurate, and not misleading. This research will also improve our understanding of the nature of potentially problematic green claims, and could help regulators prioritise their efforts to improve compliance with the law and enforce against breaches. 

We used a novel research methodology in which we deployed a large language model to analyse product descriptions on retailer websites. The model was asked to apply an assessment framework of 24 checks based on five of the six principles in the Green Claims Code. The framework was tested and developed iteratively using a pilot sample to ensure the questions were robust, and supplemented by consumer tests (the questions in the framework were asked to a panel of consumers) to compare the performance of the model with typical consumers. The use of AI allowed us to analyse a substantially larger sample of product descriptions when compared to using human review.

Our research found that from an initial sample of just over 8,800 product descriptions, more than a fifth contained at least one green claim (22%). We then analysed the green claims made in a reduced sample of 1,000 products that were representative of consumer spending patterns. We found that only 16% of these products did not fail any checks, while 62% of the products failed checks related to at least two of the Green Claims Code’s principles. 

Figure 1: Percentage of products by the number of Green Claims Code principles they failed a check for

The product categories with the highest proportion of products failing checks on multiple principles were cleaning products, electronics and accessories, and personal hygiene products.

Products are most likely to fail check relating to:

• Principle 2: “Claims must be clear and unambiguous”
  • Approximately two-thirds (65%) of products with a green claim failed at least one of the checks under this principle. The most common reason for this was that vague or general terms, such as ‘sustainable’ or ‘environmentally friendly’, are used without justification, explanation or context.

• Principle 5: “Claims must be substantiated”
  • A substantial number of products did not substantiate their claims by either providing evidence in the description, or signposting consumers to where it can be found. Nearly six out of 10 products did not provide evidence in the description, and of these products, 96% did not supply a link, QR code or reference to where consumers can find the evidence.

The research provides an estimate of the prevalence of failure to comply with specific areas of the Code amongst businesses. It indicates that there could be widespread compliance issues and that there are common pitfalls for businesses. The implication is that there is potential for regulators to provide more support to well-intentioned businesses who may be unclear about how to apply the Code, but also to conduct investigations and take enforcement action against those firms who may be deliberately failing to comply. By shedding light on the parts of the Code businesses are less likely to comply with, we hope it could help regulators prioritise future compliance and enforcement activities.

Our findings also highlight the need for research into how different compliance failures influence consumers, and their potential to cause consumer harm. This would further strengthen our understanding of which types of compliance issues should be prioritised. 

However, it is ultimately for businesses themselves to take responsibility for ensuring their environmental claims are accurate and comply with the Code. Our findings show that businesses need to be more vigilant about how their products are marketed.

Introduction

Sustainability matters to consumers and the green claims made by businesses influence consumer decisions. Over a fifth of consumers report that they are often or always influenced by information on packaging or labels about the environmental impact of a product when purchasing food (28%), electronics (24%), and household goods (21%) [1] and consumers claim they are willing to pay up to 10% more for sustainable products. This has been corroborated by experimental studies, with results showing that consumers were willing to pay more for products labelled with sustainability claims.

However, it can often be impossible for individual consumers to verify whether claims about the environmental impact of a product are genuine and accurate, and recent research has found an uncritical acceptance of environmental claims among consumers. This may incentivise some businesses to make misleading green claims, to the detriment of consumers, honest businesses and the environment. If consumers are misled into buying products that are not as good for the environment as claimed, then they may overpay. Meanwhile, genuine firms selling more sustainable products are placed at a competitive disadvantage. This destroys the financial return from being more sustainable and with it the financial incentive for businesses to act more sustainably.   

Recent regulatory actions suggest that the use of misleading green claims is not uncommon. The Competition and Markets Authority (CMA) has conducted sectoral level investigations of misleading green claims in the fashion and Fast Moving Consumer Goods (FMCG). It has published the Green Claims Code (the Code) as guidance for businesses to help them comply with the law. Manufacturers, wholesalers and retailers can all be responsible for ensuring that claims made to consumers are compliant with consumer law [2]. Meanwhile, the Advertising Standards Agency has investigated individual companies who are in breach of the UK Code of Non-broadcast Advertising and Direct & Promotional Marketing (CAP Code), resulting in several high-profile decisions, including cases involving airlines’ greenwashing practices. Meanwhile, the Advertising Standards Agency has investigated individual companies who are in breach of the UK Code of Non-broadcast Advertising and Direct & Promotional Marketing (CAP Code), resulting in several high-profile decisions, including cases involving airlines’ greenwashing practices.

However, there has been little research into the prevalence of potentially misleading green claims in the UK. In 2021, the CMA and its international counterparts estimated that 40% of green claims made globally could be misleading to consumers, but as the proliferation and regulation of green claims can differ by country, there remains a need for better understanding of the prevalence in the UK. Further, greater knowledge of the nature of green claims would inform our understanding of which aspects of green claims are more commonly problematic, indicating areas in which regulators should prioritise their efforts to increase compliance with the law and enforce against breaches.

In this study, we explore the prevalence of green claims that are less likely to comply with the Green Claims Code, and therefore are less likely to be clear and accurate, and in some cases, more likely to mislead. This was done using an innovative research methodology in which 1,000 product descriptions on retailer websites were analysed by a large language model using an assessment framework of 24 checks that we created based on the Code. This framework was tested and developed iteratively using a pilot sample and supplemented by tests with a panel of consumers. Using artificial intelligence in this way allowed us to analyse a substantially larger sample of product descriptions than we could have managed using human review. 

Our full methodology is described in Chapter 1, before the findings are presented in the following three chapters. In Chapter 2, we consider the prevalence of green claims, how the prevalence varies by product category and the different types of green claims. The extent to which green claims may not be complying with the Code is then examined in Chapter 3, and in Chapter 4, we break down this assessment by considering each of the principles of the Green Claims Code in turn. In Chapters 3 and 4, the findings from the large language model are supplemented by illustrative case studies that have all been subjected to human analysis. We then summarise our findings and the implications of our research in Chapter 5.  

Chapter 1: Methodology

To estimate the prevalence of green claims which are less likely to comply with the Green Claims Code, we focused on analysing the green claims found in product descriptions. The product descriptions were collected through online multi-brand and individual retailer websites rather than in-store to ensure we can obtain a sufficiently large sample of products within a reasonable timeframe.   

The methodology can be summarised as follows, with more detail in the rest of the chapter:

Selection of sample products and websites

To create a sample of products for analysis, we identified an initial longlist of product categories and corresponding product sub-categories, representative of consumers’ consumption patterns taken from the ONS’ Consumer Prices Index (CPI) basket of goods. From this, we then excluded product categories in service sectors or monopolistic markets where green claims were not likely to be relevant, such as domestic and household services, catering services, and education services. 

Further minor adjustments were made to the product sub-categories, with details in Annex 1. For each product sub-category, a set of common products was chosen to represent that sub-category (eg white and brown bread for bread, and full-fat and semi-skimmed milk for milk). 

To choose the websites from which we collected  product descriptions, we selected a mix of well-known third-party retailer websites that offer multiple brands, and the brand-owned websites in categories where products aren’t typically sold on multi-brand websites. Where possible, market share data from Statista and Mintel were used to inform this selection. This resulted in 32 selected websites, with nine multi-brand third-party websites and 23 individual brand websites.

Crucially, where products are sold through third-party websites, it may not be clear who wrote the product descriptions. Brands do usually provide approved marketing wording to those retailers, but retailers may make edits, and updates made by the brand can take some time to trickle through. Product descriptions used may vary across retailers and from what is found on the brand’s own website.     

Detailed product categories, sub-categories and products alongside the full list of websites that were selected can be found in Annex 1.  

Data collection

The product information was gathered in June 2024 through a combination of automated and manual data collection, capturing the following details:

• Product name;
• Product description;
• Price;
• Images attached to product such as certification(s), marketing images, and pictures of product;
• Metadata, including search terms used; and
• Links in the product description, if any

To ensure the products collected were as relevant as possible, the collection was designed to mirror the shopping behaviour of a typical consumer. A search term reflecting each product category was entered, the search was set to show the most relevant products, and the products shown on the first few pages were collected.

Some products were also manually collected where environmental claims are not product-specific or it was difficult to automate the data collection:

• Energy (electricity and gas suppliers);
• Cars, scooters, and car accessories;
• Travel (ferry fares, air fares, holiday packages); and 
• Books

Many of these manual collections were from the 23 individual brand websites. For sectors such as energy and travel, green claims were typically found on their ‘sustainability’ or ‘sustainable collections’ pages instead of on the product descriptions, so we collected the green claims from those sections. 

Creation of the final sample

We initially collected 20,000 products in total across our product categories. After deduplication and removing all instances where the product did not match the term used to search for it, there were 8,800 products left.

From this initial sample, we created a smaller sample of 1,000 products that had green claims and this sample is used for our analysis. The green claims were identified and extracted using GPT-4o (GPT), with multiple iterations made to the prompt to ensure the extracted claims were consistent and accurate [3]. A manual review was conducted on 1% of all products with and without green claims and 5% of products of the final 1,000 products to ensure the final extraction was robust [4].

This final sample of 1,000 products was designed to be broadly representative of the ONS’ CPI basket so that the number of products in each category are indicative of the share of consumer spending on that category. The full list of products and number of products collected for each can be found in Annex 1.

Development of the assessment framework

Creating the framework

The assessment framework used to identify whether the green claims complied with the Code was created based on the Code’s principles and its corresponding checklist. This framework incorporates five of the six core principles as its foundation:

• Claims must be truthful and accurate
• Claims must be clear and unambiguous
• Comparisons must be fair and meaningful
• Claims must be substantiated
• In making the claim you must consider the full life cycle of the product or service

We then developed questions in the assessment framework, which were included in prompts to GPT. The large language model also allows us to analyse a far larger sample of green claims than would be possible with human reviewers.

However, one limitation was that we could not use GPT to fully assess the product descriptions against all aspects of these principles. For example, with regard to truthfulness and accuracy, GPT is able to detect contradictions in product information, but unable to verify the factual accuracy of claims. However, verifying these would be nearly impossible even for a human reviewer, as this information would usually not be known to the public. We were also unable to assess certain questions related to the principle of substantiating green claims, such as whether evidence is available to or from others in their supply chain, as this information is likely known only to the business.  

We were unable to include the remaining principle at all:

• Claims must not omit or hide important information.

This is because after initial testing, we found that it was not possible to get GPT to accurately identify what information was missing that could be considered important to a consumer. 

Based on the principles and the accompanying checklist, an initial set of framework questions were developed as checks to assess whether the product information met the requirements of the principles. However, initial tests with GPT showed that some of the checks required additional questions in order to provide a coherent response, thus we added more framework questions to precede certain checks. For example, in a check on whether the comparison of a green claim was clear, there would be a preceding question asking if there were any comparisons made in the product description. This resulted in a total set of 49 framework questions for 26 checks before further development and testing. 

Iterative development

The framework questions were tested and developed iteratively using a pilot sample of 38 products across the different product categories to ensure the robustness of the questions and that GPT was able to accurately answer them. This was done by asking GPT to respond to the questions in the framework for the 38 products and refining the prompts based on the responses. We then analysed GPT's response to the questions and used prompt engineering to refine the questions where answers were judged to be illogical or made up.. These steps were repeated until GPT provided reasonable and consistent responses to the questions, resulting in a total of 45 questions, of which 26 are checks to assess whether the description complied with the principles. 

Consumer tests

As a further robustness check, we conducted consumer tests to compare the performance of GPT with average consumers. We asked a panel of 24 consumers, constructed using quotas for age, ethnicity and gender, to evaluate the 38 pilot sample products using the 45 questions in the framework. Each consumer assessed 3-4 products, ensuring every product had responses from two participants. This meant that in total, 76 assessments were completed for 38 products. 

Overall, we found that consumers’ responses were generally well-aligned with GPT’s responses, with at least a 75% agreement rate between the responses for 32 of the questions. Of the remaining 13 questions with a higher disagreement rate, we decided to drop one about whether the claims only focus on the positive environmental impacts whilst omitting negative impacts the framework, due to extensive inconsistencies between how GPT interpreted the question during each round of iteration as well as between each consumers’ interpretation. The remaining 12 questions were further examined. Two researchers from the team responded to these questions and reconciled the answers to provide a set of ‘correct’ answers, finding that GPT provided the ‘correct’ answer over 80% of the time for 6 of these questions and at least half of the time for the rest [5]. We also made adjustments to the prompts for three of these 12 questions. The full results can be found in Annex 2.

Final framework

The final version included 44 framework questions, split into three rounds as we found during the iteration process that this helped GPT to identify the different elements of the green claims before applying a judgement. From these questions, there were 26 checks which evaluated green claims against various aspects of the Green Claims Code. During the final analysis, we found GPT was making overly strict judgements on two of the 26 checks. For example, when asked about whether the evidence fully supports the scope and magnitude of the green claim made, GPT decided that the evidence did not support many of the green claims despite the evidence doing so from manual checks. Therefore, statistics are reported using only the other 24 checks, although not all products will be asked about all 24 checks as they may not be applicable for those products. Table 1 shows how these 24 checks are mapped to the principles of the Code. The full framework with the 44 framework questions (including the 26 checks) can be found in Annex 3. 

Table 1: Checks mapped to principles in the Green Claims Code

Data analysis

After the claims were extracted and analysed, we produced descriptive statistics on:

• The overall prevalence of  green claims followed by a breakdown by product category and types of green claims (presented in Chapter 2);
• The overall prevalence of green claims less likely to comply with the Code in the UK, followed by a breakdown by product category (Chapter 3);
• A breakdown of failed checks mapped to Code principles (Chapter 4).

Throughout the analysis, we highlight some case studies which demonstrate where green claims are unclear or less likely to comply with the Code, and in some cases, could also be misleading. We also include case studies which discuss how some products are complying with the Code. Responses from the brands and retailers mentioned in the case studies can be found in Annex 4.

Chapter 2: Prevalence of green claims

Overall prevalence of green claims

From the original sample of 8,800 products, 22% contain green claims in their product descriptions. This sample is not intended to be representative of consumer expenditure patterns [6], so it only provides a rough approximation of how widespread the use of green claims are among the products that people buy. This estimate excludes standard recycling labels (eg "Lid - recycle"), energy ratings (eg "E rating") and animal welfare claims, as these are either common information across products or imply benefits beyond sustainability.

There are substantial differences across product categories in how common it is to find green claims, see Figure 2 [7]. It is particularly prevalent in ‘Food & beverage’ product categories, where 46% of products in those categories carry green claims. There is also wider variation between categories in ‘Food & beverage’ than in other categories. ‘Non-alcoholic beverages’, ‘Fat & oil’, and ‘Proteins’ are among the categories with the most green claims whilst ‘Cereals’ and ‘Alcoholic beverages’ rank lower. Outside the ‘Food & beverage’ categories, ‘Cleaning products’ also frequently feature green claims.  

Figure 2: Percentage of products with green claims by product category

The high prevalence of green claims observed in ‘Food & beverage’ product categories is partly driven by organic claims related to fresh produce. For example, just over half of fruit and vegetables with green claims are organic ones. Organic claims in the food sector are regulated and manufacturers of organic products must be registered with a UK approved body.

We found that products typically only have a few green claims in their description, with an average of 2.9 green claims per product. This is fairly consistent across the product categories, with the average for each category ranging from two to four.  

Prevalence of green claims by type 

Within the representative sample of 1,000 products which contain green claims, green claims can be broken down into different types. Figure 3 shows that over half of products use a vague or general term, followed by claims of the product or its packaging being recyclable, and the use of recycled materials. Despite the focus on net zero and carbon neutrality in the media and by the government, we found only 5% of green claims include terms related to those concepts. 

Figure 3: Percentage of green claims by type

The types of the green claims are fairly consistent across product categories, with the use of vague or general terms being the most common type for a majority, followed by claims of the product or its packaging being recyclable. There are some notable exceptions like clothing and footwear where claims about the use of recycled materials are the most common, and in vegetables where organic claims are used the most. 

Chapter 3: Prevalence of failed checks

In this chapter we present the frequency with which products with green claims fail checks in the assessment framework, based on our final sample of 1,000 products. The large language model that we used to assess the green claims is unable to make an assessment of the potential legality of individual claims [8]. However, by assessing each product description against the assessment framework’s 24 checks that are derived from the Code, we are able to estimate proportions that are less likely to comply with the Code, which therefore are less likely to be clear and accurate, and in some cases, are more likely to mislead consumers. To make an assessment of whether any of the individual claims actually are misleading would require further investigation [9].

We present the overall distribution of products against the number of checks failed and the distribution of products against the number of principles failed, as well as the breakdown by product category.

Overall prevalence

Figure 4 shows the distribution of products against the number of checks failed. It shows that the vast majority of products (84%) that make a green claim failed at least one check in the assessment framework. Within this, the failure of some checks will be more concerning than others. However, we assume that, on average, the greater the number of failed checks the greater the likelihood that the product description is not compliant with the Code. About one in 11 products (9%) failed at least five checks.

Figure 4: Percentage of products by number of checks failed

Since some products may fail multiple checks relating to a single Code principle, we also present the percentage of products against the number of principles for which they failed at least one check (Figure 5). Again, we would be more concerned about product descriptions that fail checks across multiple Code principles as these are more likely to contain claims that don’t comply with the Code. In total, 62% of products with green claims fail checks relating to at least two of the Code’s principles.

Figure 5: Percentage of products by the number of principles they failed a check for

It is important to note that whilst products failing more checks and more principles may be less likely to comply with the Code, there are various levels of severity to which this happens. For any given check, some failures will be more problematic than others. For example, it is ultimately worse if product descriptions are not substantiating their claims because the businesses actually don’t hold any evidence to support their claims, compared to a situation where they haven’t substantiated their claims but the businesses do hold sufficient evidence and the claims are substantively true. Further, there are products which only incidentally refer to their product being sustainable, in contrast to others in which green claims are fundamental to the product’s marketing, as we will see from the case studies below. The context for a green claim is likely to impact the extent to which a claim influences a consumer choice, and hence the importance of a failed check, but there is insufficient research on this to definitively discuss how.   

Failure rates by product category

There are also differences in the rates with which products in different categories failed checks and principles. Figure 6 shows the proportion of products within each category that failed multiple principles. Cleaning products, electronics and accessories and personal care and hygiene products have the highest proportion of products failing multiple principles, and hence are the product categories where we believe there is a higher risk of green claims not complying with the Code.

Figure 6: Breakdown of products against the number of principles they failed, by product category

Some of the categories with lower prevalence are unsurprising given there were not many green claims found, such as ‘Cereals’, and ‘House appliances’ (refer to Figure 2 in Chapter 2). Meanwhile, ‘Homeware’ also had relatively few green claims, but despite this there was a high percentage of products with claims that failed more principles. Conversely, despite some fresh food and beverage categories such as ‘Proteins’, ‘Vegetables’ ,and ‘Fruit’ being among the categories with the most green claims (as shown in Figure 2 in Chapter 2), they had lower rates of claims that failed more principles, especially for ‘Fruit’. This is likely due to the prolific use of ‘organic’ in the categories, with nearly three-quarters of products in these categories describing themselves as organic having recognised organic certifications clearly displayed (71%). This follows strict rules about labelling food as ‘organic’ in the UK with certifications only given by UK approved control bodies.

Below are examples of green claims from the worst-performing product categories, including cleaning products, electronics and accessories, and personal care. These are examples from cleaning products and personal care for products in which green claims are used extensively and prominently, but where multiple failings can be observed; a tech product where a green claim comprises only a small element of the product description but is not substantiated; and a food product that makes multiple green claims and avoided failing any checks because the claim is accompanied by evidence of certification.

Case study

 Example 1: Cleaning products - Delphis floor cleaner (taken from Ocado)

This product listing contains a number of green claims, including some relating to the product as a whole, some relating to the product’s packaging, and some relating to the ingredients. 

The potential issues with the green claims found in this product description are:

• The Code states that claims must be clear and unambiguous. One of the green claims in this product listing is that the product is ‘made from highly biodegradable, renewable and sustainable ingredients’ (followed by a list of ingredients that the product does not contain). However, there isn’t an explanation of how or why the product’s ingredients are ‘highly biodegradable, renewable and sustainable’. The CMA’s guidance states that vague or general statements of environmental benefit are more likely to be misleading, and as a result our view is that this claim may not be compliant with the Code. 

• The Code also states that claims must be substantiated, and that ‘claims are less likely to mislead where the supporting evidence is publicly available and it is clear where and how consumers can verify the claims.’ The product listing does not signpost consumers to where specific information about the green claims can be found (including that the product is ‘made from highly biodegradable, renewable and sustainable ingredients’) - although it does provide a link to the EcoLabel site which provides a certain level of assurance relating to the overall green credentials of the product. However, this site itself doesn’t contain evidence or information relating to the specific claims about the ingredients’ biodegradability, renewability and sustainability. In the absence of links or references to supporting evidence, we believe this example is more likely to mislead compared to one where that information is signposted.

Case study

 Example 2: Personal care & hygiene - KIT & KIN hypoallergenic eco nappy pants (taken from Ocado)

 The potential issues with the green claims found in this product description are:

• Terms like "oxo-biodegradable material", "bio-based SAP", and "mass balance" are technical terms which are not explained. One of the Code’s principles is that ‘claims must be clear and unambiguous’; and the CMA’s guidance states that ‘scientific or technical language should be avoided unless it is easily understood by the average consumer’, which we don’t think is the case here. In our view, the use of such terms in these claims is not compliant with the Code.

• Some of the green claims contain comparisons eg ‘Our improved specification makes us the most accredited eco nappy on the market’ and ‘We use sustainable, plant-derived materials in contrast to most standard nappy pants, which are made from oil-based plastics’. ‘One of the Code’s principles is that ‘comparisons must be fair and meaningful’. The guidance states that comparative claims should indicate how the information that forms the basis of the comparison can be accessed in order for the comparison to be verified; however, no information is provided about how consumers can assess these claims for themselves. In our view these claims are more likely to mislead consumers than ones where that information is provided. 

• In relation to claims such as ‘produced in a carbon neutral factory’, the principle that ‘claims must be substantiated’ is relevant. The guidance states that ‘Claims are less likely to mislead where the supporting evidence is publicly available and it is clear where and how consumers can verify the claims.’ There is no signposting or links to further evidence for this claim, so in our view this claim is more likely to mislead compared to one where that information is signposted.

Case study

Example 3: Apple iPhone 15 Pro Max 5G 256GB (taken from Argos)

The potential issues with the green claim found in this product description are:

• It claims the product uses ‘more recycled materials’ but it is unclear what the basis of comparison is (eg with a previous model or other smartphones) nor was there any further explanation provided for that. One of the Code’s principles is that ‘comparisons must be fair and meaningful’; as it is not clear what the comparison is here, our view is that this claim is not compliant with the Code.

• It did not provide any evidence to substantiate this claim nor did it provide further links. The Code states that claims must be substantiated, and that ‘claims are less likely to mislead where the supporting evidence is publicly available and it is clear where and how consumers can verify the claims.’ The product listing does not signpost consumers to where specific information about the green claim can be found. In the absence of links or references to supporting evidence, we believe this example is more likely to mislead compared to one where that information is signposted.

Case study

Example 4: Fruits - Ocado Organic Tomatoes on the Vine

There are three green claims and certifications in the product description, all related to the tomatoes being organic. It passes the checks because it clearly states the certification body and number as well as the Soil Association logo, which is a widely recognised body. The evidence to support the claim that the product is organic is therefore signposted for consumers.

 Chapter 4: Findings by Green Claims Code Principle

In this chapter we explore in detail how well claims perform against each principle of the Code by analysing the failure rate for each individual check. Firstly though, we present the rate at which products failed at least one check for each principle.

As noted earlier, it was infeasible to assess the principle ‘claims must not omit or hide important information’. Table 2 provides the overall failure rate for each of the other five principles. 

Table 2: Percentage of products which failed any check for each principle

The highest rate is for the principle that states comparisons must be fair and meaningful, but the number of products that had claims which made comparisons was small, only 85 in our sample. 

In absolute terms, far more products failed checks relating to claims being clear and unambiguous, or claims being substantiated. As we explain below, the main reasons for claims not being clear and unambiguous are that: vague or general terms are not justified, explained, or evidenced; visual elements are not clearly explained, and that technical terms unlikely to be understood by the typical consumer are not explained. For claims that have a lack of substantiation the most common problem is a failure to provide weblinks, QR codes or references.

Principle 1: Claims must be truthful and accurate

For claims to be considered truthful and accurate, this principle requires claims to contain correct and true information and not to overstate environmental benefits. It also specifies that if there are conditions or caveats which apply, they are clearly set out, easily understood, and do not contradict the claim. Some of these aspects can be assessed by GPT using the assessment framework, but other aspects would be impossible to verify without further testing or obtaining more information from manufacturers [10]. The assessment framework included six checks relating to this principle. 

The percentage of products that failed each of the checks for this principle is presented in Table 3. Overall, about one in five products failed at least one of the checks under this principle (21%). This is lower than some other principles, but perhaps reflects our ability to only partially check some parts of this principle. 

Table 3: Percentage of products failing checks under Code Principle 1 - “Claims must be truthful and accurate”

All of the 1,000 products were checked to see if components not typically used or simply meeting industry standards were presented as a benefit. More than one in six products failed this check. Examples of failing products we found include plastic bottles being specified as being 100% recyclable, which implies a particular environmental benefit despite nearly all plastic bottles being fully recyclable, and ‘plastic-free’ wipes when many of the wipes being sold are already plastic-free due to upcoming legislation to ban them.

The check about displaying the percentage of organic material or an organic certification has the highest failure rate at 29%, although this relates to only 115 products which make an organic claim. Whilst some food and beverage products fail this check due to the absence of a specific organic certification, this sector is subject to strict regulations about the use of organic claims. Most failures are from non-food and beverage products, particularly textiles, where there is no such regulation.

The Code states that “consumers are likely to understand the term to mean that a product labelled as organic has a very high level of organic components. A claim that a product is organic where it falls short of this level is likely to be misleading.” Failing this check does not necessarily mean a claim is misleading or non compliant with the Code, but in our view such claims are more likely to be problematic compared to claims that do provide such information, due to the absence of that further detail or certification. Further investigation is needed in order to draw conclusions about the organic claims. However, businesses can reduce the likelihood of making a non-compliant claim as there are organic certifications such as The Global Organic Textile Standard (GOTS) and the Organic Content Standard (OCS) that brands can obtain and display. 

An example of a product description that did not pass our organic certification check is shown below.

Case study

Example 1: Seasalt Cornwall Men’s Blazer (taken from Next)

There is a green claim that this cotton blazer is organic, but it has no recognised certifications and no information about the percentage of cotton that is organic. This is unsurprising as the use of the term ‘organic’ is not regulated in a similar way to the food and beverage sector. Although this fact alone does not mean the claim is not compliant with the Code, we think that this is more likely to be problematic compared to claims that provide more information about their organic material.

Principle 2: Claims must be clear and unambiguous

This principle is about ensuring the terms and their meanings should be clear to consumers. Among its requirements are that vague terms, like ‘eco-friendly’ or ‘sustainable’, and technical terms are sufficiently explained. 

Overall, nearly two-thirds of products failed at least one of the checks under this principle (65%). However, the reasons for this vary as there are high failure rates across a number of checks, but some checks only apply to a minority of products, see Table 4. Only one check applied to almost all products, which is that it is clear which part of the product the claim relates to and this was failed in relatively few cases.

Table 4: Percentage of products failing checks under Code Principle 2 - “Claims must be clear and unambiguous”

The most commonly failed check relates to vague or general terms being used, but not explained or accompanied with more context. This was failed by 88% of the products it was applied to. According to the Code, general claims “such as ‘environmentally friendly’, ‘eco’, or ‘sustainable’ don’t provide any real indication of what is meant”. Hence, businesses should use more specific terms or include explanations. Examples of how businesses are using vague or general terms and how they can be clearer are shown below. 

Case study

Products which use vague / general terms without explaining them or using more specific terms

Example 1: Grind House Blend Ground Coffee (taken from Tesco)

In this example, there is no explanation about what ethically sourcing coffee means or what makes the farms sustainable so these appear to be vague and general statements of environmental benefit. According to the CMA’s guidance, such statements are more likely to be misleading. In our view, this claim is not compliant with the Code.

Example 2: Hey Duggee Bean Bag Duggee Soft Toy (taken from Argos)

This product information uses a vague term, “environmentally friendly”, without being clear about what makes the filling environmentally friendly, and is therefore more likely to be misleading. In our view, this claim is not compliant with the Code.

Product which avoids vague / general terms and is specific with a green claim

Example: Birra Moretti 18x330ml Lager (taken from Ocado)

The green claim presented in the product description does not use any vague / general terms. Instead, it is clear and specific about the environmental benefit of the brand. 

Within this principle, the CMA’s guidance also states that claims of products being biodegradable can be problematic as they often only apply “in certain conditions” like special industrial compost facilities rather than in people’s own homes. If no conditions apply to the claim (if, for example, the product is biodegradable through the typical methods for disposing of the product) the CMA’s guidance does not say that this must be stated, so not including information about conditions does not necessarily mean that the claim is misleading or non-compliant with the Code. However, in our view, claims lacking that further information are more likely to be problematic than claims where this is explained. We found that for nearly three-quarters (73%) of products with compostability or biodegradability claims it was unclear whether these types of caveats applied to their product.

Case study

Product which does not include caveats on biodegradability / compostability

Example: Fred & Flo Biodegradable Wipes (taken from Tesco)

There is a green claim about the wipes being biodegradable and the term is used in the name of the product itself. Whilst the wipes may actually biodegrade without requiring specific conditions, in our view it should be made clear whether any conditions do or do not apply. Further investigation would be needed in order to draw conclusions about whether the product is biodegradable under normal household conditions or requires specialist equipment, which is beyond the capability of an average consumer.

Product which includes caveats on biodegradability / compostability

Example: Dettol Cleansing Surface Wipes (taken from Tesco)

The product description is very clear that the wipes biodegrade in active landfill and composting conditions, alongside listing the test standards of the biodegradability test.

Another aspect of the principle also talks about presenting information that is relevant to the claim being made and presented in a clear way. We think this includes visual elements such as logos and badges that portray an environmental benefit but are then neither explained in the description nor a recognisable certification. We found that 70% of products with a logo or badge have one which is not a recognisable certification and do not explain it further. Examples of this include businesses creating their own badges for products without explaining what the badges are and why the badge means the product is environmentally friendly in the product description. 

The Code guidance for this principle also specifies avoiding scientific or technical terms unless they are easily understood by the average consumer, and we can see over two-thirds of product descriptions that use technical terms do not use ones easily understood and do not explain them (68%). The guidance advises that businesses should use words and phrases in their ordinary meaning and in the way that a consumer will likely understand. If they use a technical term, they should define and explain it. Examples of how scientific or technical terms are used with and without explanations are given below.

Case study

Product which uses scientific or technical terms which may not be understood by an average consumer and are not explained

Example: FLO Bamboo Sanitary Day Pads (taken from Ocado)

It seems unlikely that an average consumer will understand what is meant by PLA (polylactic acid) and SAP (superabsorbent polymer) or that they are a type of plastic material. Also, no further explanations are given about what it actually is or its ordinary meaning. In our view, this claim is not compliant with the Code.

Product which uses scientific terms which may be understood by an average consumer and are explained

Example: Ocado olive oil (taken from Ocado)

In this example, the technical term, ‘polyethylene terephthalate (PET)’ is used but there is enough information provided such that consumers will understand the key points, which are that PET is a type of plastic and the bottles are fully recyclable.

Principle 4: Comparisons must be fair and meaningful

When products make comparisons in their green claims, these comparisons need to be on a similar basis and should ‘indicate how the information that forms the basis of the comparison can be accessed to be verified’. It is relatively uncommon for green claims to include comparisons, so only 85 products were subject to the checks for this principle. It is relatively uncommon for green claims to include comparisons, so only 85 products were subject to the checks for this principle.

We found that 85% of products with a comparative green claim failed at least one of the four checks relating to this principle, as shown in Table 5. This is primarily because they do not provide any evidence to support the comparisons they make. For example, if a claim such as the "UK’s most sustainable cleaning solution" is made, there should be some explanation or information about how the evidence to support this claim can be accessed. 

The next most frequent problem was that the basis of the comparison was not made clear. More than a third (36%) do not state what the comparison is against. Some examples are provided below to illustrate these points.

Table 5: Percentage of products failing checks under the Code Principle 4 - “Comparisons must be fair and meaningful"

Case study

Products which do not make the basis of their comparison clear and provide no indication on how to access the information to verify it

Example 1: Chad Valley Wooden Puzzles (taken from Argos)

This makes a relative claim that wooden toys are “much more kind to the environment”, but it is unclear what specific environmental aspects are being compared (eg carbon footprint, resource use etc) that makes a wooden toy better for the environment nor what it is kinder than (eg plastic toys). The CMA’s guidance says that claims should make clear to consumers what is being compared and how the comparison has been made. As this claim does not, we think it is more likely to be misleading compared to claims where this is explained, and is not compliant with the Code. Additionally, it does not provide any signposting to where consumers can check the claim to support the comparison.

Example 2: Charlie Bingham Lasagne for 1 (taken from Ocado)

This product description made a claim about using “30% less cardboard!” as it’s implying a comparison but it’s not clear what it’s comparing to, whether to a specific previous product or other microwave lasagnes. As the CMA’s guidance says that claims must make clear against what or whom the comparison is made, we think this claim may not be compliant with the Code. When we checked, it also did not provide any explanation to support this claim at the link provided, such as details on cardboard use of previous packaging or other lasagnes. Comparative claims should indicate how the information that forms the basis of the comparison can be accessed in order for the comparison to be verified for accuracy.

Product which does make the basis of their comparison clear and provides evidence

Example: Kellogg's Special K Cereal Bar (taken from Ocado)

The claim about less packaging clearly identifies that it is compared to a previous pack with the same weight, and provides the specific measurements to support the claim. Information also appears to have been provided on how information that forms the basis of the comparison can be accessed, so that consumers can verify it - ie through the link to the Kelloggs website.

Principle 5: Claims must be substantiated

The principle that ‘Claims must be substantiated’ requires businesses to hold ‘robust, credible, relevant and up to date evidence’ to back their claims up. A relevant point from this principle is that ’claims are less likely to mislead where the supporting evidence is publicly available and it is clear where and how consumers can verify the claims’. There is a larger question about whether this type of signposting is helpful for consumers to make an informed decision, especially on product labels.

We found that many of the products which failed Principle 2 also failed Principle 5, as green claims which are vaguely worded tend to not be accompanied by further evidence backing their claims up. 

Over three in five products failed at least one of the checks developed for this principle, as shown in Table 6. Whilst most of the 431 products that support their claims with evidence have credible and updated evidence, many of the remaining 569 products without evidence do not indicate where such evidence can be found (eg via links,  QR codes or references). 

Table 6: Percentage of products which fail the checks under the Green Claims Code Principle - “Claims must be substantiated”

The examples illustrating some of these issues with green claims can be found below. 

Case study

Products with no evidence for the claims in the product information and no signposting to this evidence

Example 1: Mr Organic Canned Chickpeas (taken from Ocado)

There are multiple green claims here relating to the chickpeas being organic and “zero air miles” - claiming the product has not been flown by plane to reach the store. Whilst it does provide a certification for its organic claim, it does not provide any evidence or justification for its other claim of “zero air miles”. It then does not provide any further links or references for consumers to check the evidence if they wanted to. Therefore, in our view, this example is more likely to mislead compared to one where links or references are provided.

Example 2: Chivas Regal XV 15 Year Old Blended Scotch Whisky (taken from Ocado)

There is a green claim that all their sites use 100% green energy and recycle 100% of their food waste. However, there is no evidence provided and no links to further information. These are absolute claims and such definitive claims are likely to require a high level of substantiating evidence. Again, in the absence of links or references to supporting evidence, in our view this example is more likely to mislead compared to one where that information is signposted.

Product with no direct evidence in the product information but has links to further evidence

Example: Next Men’s Sweater

There is one green claim about their better cotton initiative. Whilst it does not provide specific evidence why Better Cotton is ‘protecting and restoring the environment’, it does provide a link to the initiative for consumers to read about it. 

Principle 6: Claims must consider the full life cycle of the product or service

The principle requires businesses to consider the overall impact of the product’s lifecycle and be clear about which part of the product lifecycle the claim relates to. Additionally, the Code guidance states that claims which focus only on aspects of a product that are less significant in terms of the overall environmental impact are more likely to be problematic. As this is difficult to verify, our check with regard to this is only relevant for clothing products.

Results from these checks found that a majority of products are clear about which part of the product lifecycle the claims relate to, with only 12% of products failing at least one of the two checks, as shown in Table 7. 

Table 7: Percentage of products which fail the checks under the Green Claims Code Principle - “Claims must consider the full life cycle of the product or service”

There is an additional check specifically for the clothing categories, where businesses may state that a piece of clothing is better for the environment only because they are using recycled nylon or polyester. This check was developed having taken into account comments from the European Commission in its proposal for a ‘Green Claims Directive’ where it notes the risk of misleading consumers with environmental claims about the use of recycled polyester from plastic bottles ‘if the use of this recycled polymer competes with the closed-loop recycling system for food contact materials which is considered more beneficial from the perspective of circularity’. A report from the Changing Markets Foundation also states that recycled polyester is almost always made from plastic bottles, which cannot be effectively recycled back and is likely to still end up in a landfill at the end of the product’s life cycle. The Code states that a claim could be misleading if it “suggests a product is greener than it is by ignoring some other aspect of its life cycle”. Our findings reveal that 6 out of ten relevant products fail this check, with an example of this below.

Case study

Clothing product which claims a related benefit and ignores the environmental impacts of recycled nylon / polyester

Example: Adidas ⅞ Leggings (taken from Next)

The green claims shown here are about the product being made with recycled materials, and being part of their goal to end plastic waste. The claim mentions using "85% Recycled polyester" which, despite being recycled, can still have negative environmental impacts. The use of recycled polyester, while reducing the need for virgin materials, can still contribute to the release of microplastics but that is not mentioned. Neither is the fact that recycled polyester is not biodegradable if the clothing ends up in a landfill (European Parliament research has stated that 87% of used clothes are incinerated or landfilled). In addition, there are concerns about the use of recycled polyester, which may be made primarily from plastic bottles, disrupting the bottle-to-bottle recycling loop, as clothing produced from these bottles cannot be effectively recycled back into the same quality material. As these points about other stages of the product’s lifecycle are not mentioned, in our view this claim is more likely to be misleading than one where that information is included, and may not comply with the Code. 

Chapter 5: Conclusion

Using a novel methodology to analyse the environmental claims made about products, this report sheds light on the extent to which these claims are likely to comply with the CMA’s Green Claims Code - the six principles businesses need to follow to ensure the environmental claims they make are accurate.

From a sample of about 8,800 products we found that more than a fifth (22%) had at least one green claim. This sample was reduced to a final sample of 1,000 products with green claims that was representative of consumer spending patterns. The final sample was analysed by a large language model using an assessment framework of 24 checks that was created based on the Code.  

The vast majority of products (84%) that make a green claim failed at least one check in the assessment framework. A single fail may not be cause for concern, but 62% of products failed checks in relation to at least two of the Code’s principles. Cleaning products, electronics and accessories, and personal hygiene products have the highest proportion of products failing checks for multiple principles.

Products were most likely to fail checks relating to Principle 2 that states claims must be clear and unambiguous. A common issue was that product information contained vague or general terms, such as ‘eco-friendly’ or ‘sustainable’, without these terms being explained or evidenced. Over half of products used a vague or general term and in 88% of these cases there was no attempt to explain or evidence its use. Other reasons for failing checks relating to this principle were that visual elements of the product information were not clearly explained and that technical terms unlikely to be understood by the typical consumer were not explained. 

A large number of products which failed Principle 2 also failed to substantiate their claims in the product information provided or by signposting to where evidence could be found (Principle 5). For 55% of the sample of 1,000 products there was a claim without evidence and a failure to provide a link, QR code or reference for it. 

Other less common, but still reasonably frequent issues were: firms claiming as a benefit the omission of a component that would not typically be used or simply meeting an industry standard (165 of 1,000 products); evidence for a claim not being from a credible or recognised source (74 of 431 products); the failure to provide evidence to support a comparison (73 of 85 products); and not placing clear caveats on compostability or biodegradability (64 of 88 products).

The research evidences the difficulty that consumers face when trying to verify whether green claims are genuine or accurate and it emphasises the importance of regulatory action to increase compliance with the Code. The widespread prevalence with which we observe green claims that may not comply with aspects of the Code suggests that businesses face common challenges in complying with the guidance. There is an opportunity for regulators to provide more support to well-intentioned businesses who may be unclear about how to apply the Code, but also conduct further investigations and take action against those whose actions are egregious.    

In shining a light on which elements of the Code have the greatest compliance issues, we hope this research may help prioritise future compliance and enforcement activities by regulators. However, the findings also highlight the need for further research to understand the extent to which failures to comply with the Code could lead to consumer harm. Some types of green claims may be more or less likely to influence consumer transactions and a failure to comply with the Code may cause different levels of harm depending on this. For example, to what extent do consumers respond to vague and general claims? Does technical language make a product seem more sustainable? An understanding of these issues would help to determine whether priority should be given to making sure that businesses explain and substantiate general terms or to ensure that technical terms are adequately explained.

Finally, the research shows the potential of using novel research methods to explore issues of consumer protection. Using a large language model to apply an assessment framework allowed us to analyse a far greater number of green claims than if we had needed to rely on manual analysis. The assessment of the model does need to be complemented by human review to make judgements on whether any individual claims may not be compliant with the Code. However, the large sample has allowed us to detect themes in the data and to identify where further investigations should be prioritised.

Footnotes

  [1] Based on a nationally representative poll of 2,000 consumers conducted by Yonder on behalf of Which? In 2024. ↑

  [2] The CMA’s guidance states that retailers who sell products featuring misleading environmental claims made by manufacturers or wholesalers on their packaging can be liable for those claims (as well as the manufacturer or wholesaler) and retailers should assure themselves that any such claims are accurate and not misleading. ↑

[3] This was tested on a sample of 100 products with five iterations and only one green claim had a response that changed once over the five iterations. No green claims manually found in this sample were missed by the model. ↑

[4] Results from this review found that GPT was 98% accurate on 1% of the sample and 82% on the 5% sample. The most common issue was GPT identifying claims that should not have been included so a search was done to look for the specific terms related to those claims and the products were recoded. ↑

[5] For questions where GPT made the correct assessment half of the time, this was due to small sample sizes, meaning questions were only relevant for two to four or the 38 products. ↑

[6] As described in Chapter 1, we excluded some product categories in which green claims were unlikely and over-sampled others that have a relatively high incidence of green claims. Also, the number of products in each category has not been adjusted for CPI weights. This is unlike the sample of 1,000 products with green claims on which we conduct the main analysis. ↑

[7] The product categories, ‘Bills’, and ‘Transport, and the subcategory of ‘Holiday packages’ have not been included here because, as mentioned in the methodology, the green claims in these cases were not included in the ‘product description’ but in the sustainability pages or on the cover of their sustainable ferries / holiday package sites. Their inclusion would mean 100% of these categories will have green claims. ↑

[8] This is in reference to, for example, the now revoked Consumer Protection from Unfair Trading Regulations 2008 or the provisions that have replaced them in the Digital Markets, Competition and Consumers Act 2024. ↑

[9] Ultimately, only the Competition and Markets Authority and the courts can decide whether a claim is misleading under consumer law, and whether it is unlawful may also depend on how it influences the transactional behaviour of consumers. ↑

 [10] For example, in 2023 Which? testing found that some toilet roll products which manufacturers claimed to be made of 100% bamboo (which some manufacturers presented as an environmental benefit) in fact comprised substantial proportions of wood. Such inaccurate claims could only be detected by testing the products. ↑