Episode 21: Product Data Regulation, Part 1: Green Claims and the Empowering Consumers Directive

There is a deadline coming that most product data teams have not looked at, and it has the shape of GDPR about it. Everyone assumes it will be fine, right up until the fortnight before.

From 27 September, the EU's Empowering Consumers for the Green Transition Directive applies. In short: you cannot make generic environmental claims, the green, eco-friendly, sustainable sort, without evidence to substantiate them. Penalties run to 4% of annual turnover in the member states concerned.

Claims like these have always lived with marketing, because they read like copy. They are not copy. The moment you have to substantiate one on request, it becomes a data problem, and most businesses do not hold the evidence.

The legal interpretation of your specific claims is a question for your counsel. What follows is the product data work, which you can start regardless.

One point of confusion worth clearing up

A lot of teams stopped worrying about this last year, and there is a reason for that.

The Green Claims Directive, which generated a flurry of activity when it was proposed, is a separate piece of legislation. It never became law. The Commission announced its intention to withdraw the proposal in June 2025, the final negotiations were cancelled, and its status has been in limbo since.

That withdrawal is what took the pressure off. It should not have. The Empowering Consumers Directive was adopted separately, back in 2024, and it applies from September regardless of what happens to the other one. Different law, same obligation, real date.

It also reaches backwards. The rules apply to products already on the market, not just to whatever you publish from September onwards, which puts your entire historic catalogue in scope rather than your forward content plan.

You probably do not know what you have already claimed

Here is the awkward part. Most businesses cannot tell you whether they have made these claims at all.

Three kinds of claim are worth hunting for specifically. Blanket statements of the green, eco or sustainable variety with nothing behind them. Product level climate neutral claims that rest on offsetting, which are prohibited outright. And sustainability labels or badges that no independent third party is monitoring.

Ask yourself how often anybody goes back through the historic catalogue and reads every description. The realistic answer is never. So the claims sit scattered across thousands of product descriptions, written by a lot of different people over a lot of years, with no record anywhere of where any of them came from or what evidence sat behind them at the time.

It is also not confined to descriptions. The same wording turns up in attribute values and on the badges and icons rendered on product pages, which are easy to forget precisely because nobody thinks of them as content.

Start with an inventory

The first job is unglamorous: work out what your product page content actually says and where all of it is stored. In most businesses it is not in one place.

The logical starting point is the ecommerce site, since that is where content gets rendered to customers. But that is only the first surface.

You may also be pushing green claims downstream into marketplaces and channel feeds, where you have less visibility and no easy way to recall them. And you may be making them in store, on shelf labels, without anybody in the digital team knowing it happened.

So the inventory is really a reverse engineering exercise: mapping every place a customer encounters a claim you made. It is a substantial piece of work, and it is the piece everything else depends on.

Two ways to fix what you find

Once you can see the claims, there are only two honest responses to each one.

Remove the claim. This is what one customer did, at scale. They built a dictionary of the words and phrases that could constitute an unsubstantiated claim, ran the entire content estate against it using AI, stripped the claims out, and repurposed the copy without them. Crucially, the pages did not get worse. You can showcase a product's features and benefits perfectly well without making a direct environmental claim about it.

Replace it with a verifiable fact. Where you do hold verified information, use that instead. Ninety-three per cent recycled aluminium is a specific, checkable, defensible statement about a product. Eco-friendly is a broad claim that invites a question you may not be able to answer. The specific version is better marketing anyway.

The retailer we worked through this with had thousands upon thousands of products, and it was never one description each. Titles, descriptions, SEO content, all carrying the same exposure. They were a high street name, which is to say highly visible, which is exactly the profile a regulator makes an example of.

AI speeds this up enormously. It does not make it small.

Then stop it happening again

A retrospective clean-up that is not followed by a process change is a clean-up you will repeat in six months.

The obvious trap is content generation. Plenty of businesses are now writing descriptions continuously with AI, and almost none of them have trained those models to avoid this vocabulary. So the same claims get regenerated faster than anybody can strip them out, and the team ends up permanently retraining tone of voice guidelines instead of fixing the rule.

What that needs is gating. Content enrichment should not be able to publish an environmental claim unless a verified baseline exists to support it. That is a governance control, sitting in the process rather than in somebody's judgement.

Which is where these conversations always end up. Get the data right upstream, hold trusted verifiable sources, and build content production on top of those rather than alongside them.

The takeaway

The reason this catches businesses out is that it does not look like a product data problem until the day somebody asks you to prove something.

At that point the question is not whether the claim was reasonable when it was written. It is whether you can retrieve the evidence, for that product, on request. Everything upstream of that moment is either an asset or a liability, and most catalogues currently hold rather more of the second.

So run the inventory now, while the choice of how to respond is still yours. Strip out what you cannot support, replace what you can with something specific, and put a gate in front of whatever writes your content next.

The date does not move, and being visible is not a defence. It is the opposite.

Listen to the full episode

Episode 21 of Product Data Weekly is available now, the first in a three part series on product data and regulation. For more episodes and the weekly newsletter on operational issues inside product data and ecommerce teams, visit productdataweekly.com.