AI Footprint Tracker

Local, scoped estimates of AI conversation tokens, electricity, carbon and water.

See the electricity, carbon and water behind every AI chat. AI Footprint Tracker adds a small bar under the message box of ChatGPT, Claude, Gemini, Perplexity and NotebookLM. After each reply it shows the tokens read and written and an estimate of the electricity, carbon and water the reply needed. Open Details for the full picture of the chat. WHAT YOU GET • A compact bar under each chat: requests, tokens in and out, electricity, carbon and water. • Details: the figures for the open chat, everyday comparisons (LED bulb minutes, phone charges, distance by car, tablespoons of water), how the same answer would compare on a smaller model, and how each figure was estimated. • A popup with your footprint for today, the last 7 days or the last 30 days, split by AI tool. • A dashboard with your tools and models, a monthly recap and your settings, in light or dark mode. • Images, speech and video are estimated too, with the method shown in Details. HOW IT ESTIMATES The tracker counts the text you and the AI write, finds the model that answered, and uses EcoLogits, an open-source method, to calculate electricity, carbon and water. EcoLogits gives a range for each figure. By default the tracker shows the Midpoint of that range; you can choose Minimum or Maximum in Settings. These are estimates, not measurements from the AI providers. The full methodology, with every equation, is built into the extension and published at aifootprinttracker.com. PRIVATE BY DESIGN • Personal use needs no account. Counting happens in your browser and the records stay in your browser. • The text of your chats is processed only in memory to count it. It is never stored and never sent anywhere. • No analytics, no advertising, no tracking. FOR ORGANIZATIONS (OPTIONAL) An organization can register for free at aifootprinttracker.com (up to 50 members) or run its own server. Members connect their browser in the extension with their work email and a code sent to it. From then on only numbers are sent to the organization: site, model, token counts, times and estimates, never chat text or titles. Administrators see their organization's totals and can share them with all members. In a small team, shared totals can show one person's usage. OPEN SOURCE MIT licensed. Source code: github.com/pranitnale/AI_Footprint_Tracker This independent project is not affiliated with or endorsed by OpenAI, Anthropic, Google or Perplexity.