Your Sustainability Claim Now Gets Rebuilt by a Machine Before Anyone Reads It

For most of the last two years, the conversation about AI and content has been about generation. Can it write the post, draft the report, fill the page; but that was always the less interesting half.

The half that matters now is what happens on the way in, not the way out. When someone asks an AI assistant whether your company is any good on sustainability, it doesn't hand them your carefully worded statement. It reads what it can find about you, breaks it into parts, decides what it thinks is true, and rebuilds an answer in its own words. Your claim goes in. Something reassembled comes out. You're not in the room when it happens.

If you think the likes of Claude, ChatGPT, Gemini and Perplexity simply read your webpage like a person, you're thinking about it all wrong, and you're losing.

People have started calling the work of influencing that process Generative Engine Optimisation (GEO). It's being sold as the new SEO, the next thing to game. I think that framing misses what's actually changed.

Here's what I've been noticing: a machine can only reconstruct a claim accurately if the claim was built to be taken apart in the first place.

Give it "we're committed to a more sustainable future" and there's nothing to reconstruct. No number, no boundary, no evidence, no date. The model either drops it as filler or, worse, fills the gaps itself with whatever the rest of the internet says companies like yours usually mean. You've handed your reputation to an averaging machine.

Now give it something specific. "We cut packaging weight by 18% in 2025, measured against the previous year." There's a structure to that. A figure, a scope, a timeframe, a method. The model can lift it, check it, and carry it into an answer with the meaning intact. The claim survives the journey because it was already broken into the parts a machine needs.

I've watched this happen up close, building an AI tool that reads sustainability claims for a living. The pattern is consistent. Vague claims don't get rewarded for sounding ambitious. They get hollowed out. The careful, specific, evidenced ones are the ones that come out the other side resembling what you wrote.

Which is a strange thing to sit with, because it's the same argument we've been making for years. Only the stakes have moved.

Clarity used to be about trust. A reader who understood your claim was more likely to believe it. Substantiation used to be about compliance. A claim you could prove was a claim regulators couldn't touch. Both still true. But there's now a third reason, quieter and more mechanical than either. A claim you can break into parts is a claim a machine can carry accurately when you're not there to explain it.

The companies that struggle with this won't be the ones making bold claims. They'll be the ones making soft ones. The "proud to be on a journey" school of sustainability writing was always weak. Now it's also invisible, because there's nothing in it for a machine to hold onto.

None of this needs a new playbook. It needs the discipline most sustainability communication still avoids. Say what you did. Put a number on it. Name the boundary. Show the evidence. Write the claim so that someone, or something, can take it apart and put it back together without losing the truth in the middle.

That's the part worth getting right. Not because there's an algorithm to please, but because the version of your claim that reaches people is increasingly one you didn't write. The least you can do is make it hard to get wrong.