AI Wattch: Seeing the Energy Cost of Your Prompts

We talk a lot about the environmental cost of AI at the infrastructure level: data centres, GPU farms, cooling systems. But there is another layer that gets far less attention, and it is one we each have direct control over.

Every prompt you send uses energy. A short, well-scoped request uses a small amount. A vague, sprawling one that produces a mediocre answer (and requires several follow-ups to fix) uses considerably more. That inefficiency is invisible to most users, and invisible problems tend not to get solved.

What AI Wattch actually does

AI Wattch is a browser extension built by Antarctica and IT Climate Ed. It sits quietly in the background while you use ChatGPT, Claude, or Gemini, and tracks the energy and carbon impact of your sessions in real time. Using something called the One-Token Methodology, it estimates compute intensity based on real-world model efficiency data and translates your activity into figures you can actually understand: energy used, CO2 emitted, and efficiency across sessions.

Crucially, it does all of this locally. Nothing is sent off your device. Your conversations remain private.

It is about awareness, not guilt

The framing from the team behind AI Wattch is worth noting. This is not a tool designed to make you feel bad about using AI. It is designed to make the invisible visible, which is a different thing entirely.

Once you can see that a vague prompt costs significantly more than a precise one (and often produces worse output), the incentive to improve your prompting habits becomes obvious. Better prompts save energy and get better results. That is a rare case where the environmentally conscious choice is also just the smarter one.

Why this matters at an organisational level

The technology sector is on course to represent around 8% of global emissions. Individual prompts are small, but organisations running AI at scale across hundreds of users start to accumulate meaningful compute waste, particularly if nobody has thought about prompt quality or model selection.

Tools like AI Wattch point to something we think about a lot at Futureformed: responsible AI use is not just a governance or compliance question. It starts with the habits of individual users. If your teams are prompting inefficiently, that is both an environmental issue and a cost and quality issue.

Worth installing

AI Wattch is free, private, and gives you a genuinely useful window into something most AI users have never thought about. If you are working with clients or colleagues on sustainable AI adoption, it is a good conversation starter too.

You can find it on the Chrome Web Store. If you are interested in how responsible AI fits into a broader transformation or adoption strategy, feel free to get in touch or book a discovery call.