Droplet: water used by your AI chats

Estimate the water used by your AI chat usage. No data collected, no network calls.

Droplet shows you how much water your AI usage actually consumes. "One ChatGPT query drinks a glass of water." You have probably seen the claim. Droplet lets you check it against your own usage instead of taking anyone's word for it. WHAT IT DOES Estimates the water behind your AI conversations Breaks it down by this chat, today, and this month Turns the number into something you can picture, like a teaspoon, a cup, or a shower HOW IT WORKS Droplet counts your assistant responses and how long they are, then applies published water-use figures for two things: the water used to cool data centers, and the water used to generate the electricity that powers them. The second one is usually the bigger share, and it varies a lot depending on how that electricity is produced. The estimate is built on two published sources: Elsworth et al. (2025), Measuring the environmental impact of delivering AI at Google Scale https://arxiv.org/abs/2508.15734 Li, Yang, Islam and Ren (2025), Making AI Less "Thirsty", Communications of the ACM https://cacm.acm.org/sustainability-and-computing/making-ai-less-thirsty/ PRIVACY Your prompts and the AI's responses are never read, stored, or sent anywhere No account, no sign-in, no tracking, no analytics Zero network requests. Nothing leaves your browser Everything is kept in local browser storage, and you can clear it at any time Don't take our word for any of it. The full source code is public: https://github.com/ThanveerulHuq/droplet SUPPORTED SITES ChatGPT today. Claude, Gemini and Grok are on the way. A NOTE ON ACCURACY Nobody can tell you exactly how much water a single prompt used. It depends on which data center served your request and how electricity is generated in that region, and neither of those is visible from your browser. Droplet gives a reasoned estimate using the best published figures available and shows its working. If you think a number is off, open an issue on GitHub.