Greenwatch: How widespread is greenwashing by UK businesses? Which? Investigates

A recent report published by the consumer body Which?, ("How green are green claims?"), considers the prevalence of potentially misleading green claims in the UK. These are claims that show how a product, service, brand or business provides a benefit or is less harmful to the environment. Which? assessed green claims against the Green Claims Code published by the UK consumer protection regulator, the Competition and Markets Authority (CMA) (see our October 2021 update).

The report highlights the continued occurrence of misleading green claims with 62% of sampled products failing checks relating to at least two or more of the Green Claims Code's principles. The report's findings therefore raise significant questions about the UK's current product regulatory and consumer protection frameworks, and whether more needs to be done to tackle greenwashing.  

Both the findings of the report and the assessment methodology applied by Which? provide useful guidance to businesses looking to assess the compliance of their claims and the risk of enforcement action.

The Which? Report

Methodology

A notable aspect of the report is the use of a structured AI-assisted methodology to undertake the assessment. This allowed Which? to "analyse a far larger sample of green claims than would have been possible with human reviewers", significantly improving the scale, efficiency and consistency of assessing the compliance risk in making green claims. In particular: 

• Product data was collected from 32 online retailers. These were a mix of multi-brand platforms and individual brand-owned websites, resulting in an initial dataset of 20,000 products, selected from categories aligned with the Office of National Statistics' CPI basket of goods.
• A final sample of 1,000 products making green claims was selected for analysis after deduplication and filtering.
• OpenAI’s GPT-4o was then used to extract environmental claims from product descriptions.
• A series of questions and prompts was developed and applied in GPT based on five of the six principles of the Green Claims Code (the "assessment framework") as follows: 

1. Claims must be truthful and accurate.

2. Claims must be clear and unambiguous.

3. Comparisons must be fair and meaningful.

4. Claims must be substantiated.

5. Claims must consider the full life cycle of the product or service. 

The Green Claims Code principle requiring businesses not to omit or hide important relevant information was excluded from the above criteria due to GPT's inability to identify what information was missing that could be considered important to a consumer. 

• The assessment framework was refined through an iterative process involving prompt engineering, logic testing on a pilot sample and validation through a 24-person consumer panel to compare the performance of the AI with average consumers. Which? found that the responses of its consumer panel were "generally well-aligned with GPT's responses, with at least a 75% agreement rate between responses for 32 of the questions". However, Which? had to manually iterate a number of questions which resulted in significant inconsistencies between how GPT interpreted the question compared to consumers’ interpretation. Unsurprisingly, the AI model was unable to verify the factual accuracy of claims requiring access to proprietary or supply chain data. It also struggled to assess omissions, for example, where material information was left out.
• The final assessment framework produced consisted of 44 questions.

Findings 

The Which? report highlights that over one fifth of consumers are influenced by the environmental impact of a product based on claims made on its packaging or labelling, and are willing to pay up to 10% more for products that are percieved to be "sustainable". Despite the clear impact that green claims have on consumers' purchasing decisions, little research has been conducted into the prevelance of potentially misleading green claims in the UK. 

The report concluded as follows: 

• Product Categories with most green claims: The highest number of green claims were found in cleaning products (55%) and the food and beverage product categories, where 46% of products in those categories carried green claims. Within the latter, green claims were found to be particularly widespread in relation to non-alcoholic beverages (73%), fat & oil (61%), proteins (55%), fruit (49%) and vegetables (48%). This partly reflected the use of 'organic' claims related to fresh produce. For example, over half of fruit and vegetables with green claims are advertised as being 'organic'.
• Types of green claims used across the Product Categories: Over half of the sampled products (56%) were found to use vague or general terms, such as "sustainable", "good for the planet" and "eco". A significant number of products also claimed that the product or its packaging was recyclable as a benefit (39%) or used recycled materials (26%). The types of green claims used were fairly consistent across product categories, with some notable exceptions, such as the clothing and footwear category, where the use of recycled materials are most common, and certain food categories, where organic claims are used the most.
• Adherence to the Green Claims Code: It was found that 84% of products failed at least one check and in total, 62% of products failed at least two or more checks against the Green Claims Code's principles. At a product category level, the highest proportion of failed checks across multiple principles was in respect of cleaning products (96%), electronics and accessories (91%) and personal hygiene products (77%). Despite 73% of non-alcoholic beverages containing green claims, only 42% failed checks across multiple principles – the lowest of the product categories sampled. A breakdown of how well claims within the representative sample performed against each principle of the Green Claims Code included in the AI assessment framework is summarised below:
  • Claims must be truthful and accurate: 21% of products failed at least one of the checks under this principle. Common issues identified included organic claims which lacked verification or clarity, and claims implying an environmental benefit for features that are already an industry standard, such as "100% recyclable" plastic bottles and "plastic free" wipes.
  • Claims must be clear and unambiguous: nearly 65% of products failed at least one of the checks under this principle. Common issues identified were in relation to vague or general terms such as "eco-friendly" or "sustainable" being used. The report also identified a lack of caveats on claims relating to composability and degradability, unexplained logos or badges and unclear or undefined technical terms.
  • Comparisons must be fair and meaningful: from the more limited sample of 85 products, 86% of products with a comparative green claim failed at least one of Which?'s four checks relating to this principle. This was primarily due to a failure to provide any evidence in support of the comparisons made, or the basis of the comparison not being clear.
  • Claims must consider the full lifecycle: only 12% of products failed at least one of the two checks under this principle. Despite this, 61% of clothing products claiming environmental benefits from recycled materials failed to consider broader negative impacts, such as microplastic pollution and non-biodegradability.
  • Claims must be substantiated: 62% of products failed at least one of the checks under this principle. Most failures involved absolute claims or no evidence and no links or references being provided to consumers to verify claims made (such as via links, QR codes or cross-references).

Conclusion 

The findings of the Which? report highlight the persistent and widespread use of green claims in the retail sector. This is despite the increased regulatory scrutiny and enhanced enforcement powers being afforded to regulators in the UK for breaches in product safety and consumer protection law, including in respect of misleading claims (see our September 2024 update). Businesses should be attuned to the common areas of potential non-compliance raised in the report to prioritise future compliance efforts.

The report further raises the potential usefulness of leveraging advanced AI technologies to efficiently identify patterns of non-compliance at scale. This may enable regulators and enforcement bodies to better prioritise investigative and remedial efforts. The methodology used by Which? demonstrates the ease with which regulators can now assess market actors at scale. It also highlights the ability for regulators to scan continuously wherever the right data exists, which facilitates ongoing monitoring. This is a practice that we have seen the Advertising Standards Authority (ASA), the UK's independent advertising regulator, utilise through its Active Ad Monitoring systems. These systems harness AI to proactively search for online ads that break UK advertising rules.

AI tools can also be used to support compliance activities. They already provide an important element in the risk assessment toolkit, and the accuracy, speed and capability of the technology will undoubtedly continue to improve. However, the identification and ultimate assessment of green claims will continue to require human-led review and expertise for the foreseeable future.

Tips for businesses 

Practical takeaways for businesses making green claims to adopt are: 

• Review the product categories, claims and bases of non-compliances focused upon in the report, given these are now likely to be squarely within the remit of enforcing regulators.
• Pay particular attention to any green claims made online which are easily assessed adopting similar mechanisms to the AI tools used by Which?.
• Reassess the Green Claims Code criteria in detail, including the guidance for the fashion industry published by the CMA (see our September 2024 update) and the ASA's guidance and rulings, and attempt to comply in full with those requirements.
• Innovate but do not over-rely on AI: combine AI with subject matter experts who can guide the AI effectively and make a determination for businesses in respect of any green claims made.

Other Authors: Miran Bahra, Associate; Freddie Freeman, Trainee Solicitor; Gohto Saikawa, Risk Advisory