Greenwashing by Chatbot: How to Spot AI-Written Sustainability Claims

Sustainability pages used to take a company weeks to write. Now they take an afternoon. Ask a chatbot for an environmental statement and you'll get three polished paragraphs about "our commitment to a greener tomorrow" and "paving the way for a more sustainable future". It reads well. It also says almost nothing. And that gap between how a claim sounds and what it proves is exactly where greenwashing lives.

Why AI makes greenwashing easier

A language model is built to produce plausible text, not true text. Ask it to describe a product's environmental benefits and it will happily supply them, whether or not they exist. It doesn't know what share of your packaging is actually recycled, what your emissions baseline was, or whether the "biodegradable" bag breaks down anywhere other than an industrial composting plant. So it reaches for the vocabulary of sustainability instead (eco-friendly, green, planet-positive, responsibly sourced, carbon neutral).

Those are precisely the words regulators are cracking down on. In the European Union, new consumer rules that began applying in September 2026 ban generic environmental claims such as "eco-friendly" or "climate friendly" unless a company can show recognized excellent environmental performance, and they ban claims that a product is climate neutral on the strength of carbon offsets. In the United States, the FTC's Green Guides have long warned that broad, unqualified environmental claims are hard to substantiate. A chatbot writing marketing copy isn't thinking about any of that.

How to spot an AI-written green claim

Big words, no numbers. Real sustainability progress is measurable: "42% recycled content by weight", "emissions down 18% against our 2022 baseline". AI copy prefers adjectives. If a page uses "sustainable" ten times and never gives a figure, be suspicious.

No scope. A "recyclable" label should say whether it covers the whole package or just the cardboard sleeve, and where it can actually be recycled. Most curbside programs still don't take soft plastics; a "100% recyclable" pouch may be recyclable only in theory.

No proof you can check. Look for third-party certifications you can verify, a published report, or a named methodology. "Certified" with no certifier named is a warning sign.

The familiar chatbot phrasing. "We are committed to making a positive impact." "Together, we can build a brighter future for generations to come." These lines appear in thousands of AI-generated statements, and they're the verbal equivalent of a stock photo of a leaf.

Promises instead of results. Statements that talk entirely about aspirations and journeys and never about what has already been done are easy for a model to write and easy for a company to publish without evidence.

A job for editors, too

News sites and trade blogs receive a steady flow of press releases and guest posts about green products, and a growing number of them are machine-written. Running suspect copy through a checker can help sort them. ScanForAI, for example, can check a pasted release or a live URL and marks the specific lines that read as machine-written, which gives an editor something concrete to send back. A high score isn't proof of anything on its own, and a human can greenwash just as well as a chatbot. But a release that is both machine-like and vague is a good candidate for a few hard questions before it goes anywhere near publication.

The questions are the same ones that work for any environmental claim: what exactly is being claimed, what the number is and what it's compared against, who verified it, and where the data can be read. A company with real results will have answers ready.

The real cost

Greenwashing is more than a marketing problem. Every empty "eco-friendly" label makes it harder for the businesses doing the real work, like the recycler with audited diversion rates or the hauler that actually sorts what it collects, to stand out. Cheap, generic sustainability copy drowns out the claims that matter. The fix is old-fashioned: specific numbers, honest limits, and evidence anyone can check.