AI Impact Tracker quantifies the environmental footprint of ChatGPT, Google AI Mode, and Google AI Overview usage during the inference stage. Using the user's coarse location, based on the country, or state or province and watershed (if in the United States or Canada), AI Impact Tracker estimates the carbon, water, and energy usage associated with that activity. Specifically, we calculate the amount of energy, water, and carbon associated with each individual query response, focusing only on the inference stage (and leaving out the energy consumed in other stages of the models' lifecycles, including their training) leading to likely (largely) underestimating the full environmental footprint of using ChatGPT and Google AI Mode/Overview. The extension collects the number of output tokens in a generated response and coarse location data and securely and anonymously sends this to a backend server that both performs these calculations and displays and analyzes aggregated usage and environmental data on this website to consider larger-scale impacts of generative AI. The goal of this project is to raise public awareness to these impacts, both at an individual level and at community-wide levels by sharing analysis of aggregated data here and privately displaying individual estimates in the extension's popup window for individual users.
| Tokens | KWh | CO2e (g) | Miles Driven | Water Consumption (mL) | Water Bottles (500 mL) | |
|---|---|---|---|---|---|---|
| Total | ||||||
| ChatGPT Total | ||||||
| Google AI Mode & Overview Total | ||||||
| Average Per User Usage Total | ||||||
| Average per Response (All Models) | ||||||
| ChatGPT Average per Response | ||||||
| Google AI Average per Response | ||||||
| Scaled Total ChatGPT Userbase Annual Usage |