Inspiration
What it does
AI tools like ChatGPT, Claude, and Gemini are becoming part of everyday life, but most people never see the physical resources behind each interaction. Every prompt requires computation, which means electricity, water for data center cooling, carbon emissions, and financial cost.
We built PromptFootprint because we wanted to make that invisible impact visible.
At the same time, we noticed another problem. People often send AI models prompts that are much longer than necessary. Extra filler words and repeated context can increase token usage without improving the answer. We realized sustainability and efficiency could be solved together.
Instead of only telling users that AI has an environmental impact, PromptFootprint gives them a practical way to reduce it.
What it does PromptFootprint is an AI efficiency and sustainability platform that measures and reduces the resource impact of AI prompts.
As users interact with tools like ChatGPT, Claude, and Gemini, PromptFootprint analyzes their usage and estimates metrics such as:
Tokens consumed Electricity usage Water consumption Carbon emissions Estimated computational cost PromptFootprint then helps users reduce those numbers through a built-in prompt optimizer. It identifies unnecessary wording, repetition, and filler, then generates a shorter version of the prompt while preserving the user's original intent.
Users can also view their impact over time through an analytics dashboard, making AI consumption something they can actually understand and improve.
Our goal is not to make people use AI less. It is to help people use AI better.
How we built it PromptFootprint combines a browser-based interface, a Chrome extension, and a backend system that tracks and analyzes AI usage.
The extension detects interactions with major AI platforms and estimates token consumption based on the user's prompts and model usage. We created a resource-conversion framework that translates those tokens into understandable estimates for energy, water, carbon emissions, and cost.
We also developed the Prompt Optimizer, which analyzes prompts and removes unnecessary text while attempting to preserve meaning, instructions, constraints, and context.
The platform then aggregates this information into a dashboard where users can track their AI footprint across sessions.
We designed the experience to feel less like a carbon calculator and more like a tool people would actually want to keep installed. The environmental statistics are important, but saving tokens, reducing costs, and improving prompt efficiency gives users an immediate reason to use the product.
Challenges we faced One of our biggest challenges was that there is no universal measurement for the environmental impact of a single AI prompt.
Different models require different amounts of computation, and companies do not publicly disclose every detail about their infrastructure. Energy consumption can also vary based on model size, hardware, data center location, cooling systems, and response length.
Instead of pretending that every number could be perfectly exact, we focused on building a transparent estimation framework using available research and model-specific assumptions.
Another major challenge was prompt optimization. Simply shortening a prompt is easy. Shortening it without removing important information is much harder. We had to balance token reduction with response quality so that users would not save resources at the expense of getting a worse answer.
We also had to account for differences between AI platforms, changing interfaces, uploaded files, model selection, and the many different ways users interact with AI.
What we learned Building PromptFootprint changed the way we think about AI efficiency.
We originally approached the problem primarily from an environmental perspective, but we learned that sustainability becomes much more powerful when it aligns with something users already care about.
Reducing tokens can mean:
Lower environmental impact Lower API and inference costs Faster interactions Less unnecessary context More efficient AI systems We also learned how difficult it is to translate academic research into something understandable for everyday users. Numbers like kilowatt-hours or liters per inference do not mean much on their own. Designing PromptFootprint forced us to think carefully about how data should be calculated, communicated, and visualized.
Most importantly, we learned that small improvements can become significant at scale. A few unnecessary tokens may seem insignificant in one conversation, but when billions of AI interactions happen every day, efficiency matters.
What's next We want PromptFootprint to become a complete efficiency layer for everyday AI usage.
Future versions will improve model detection, expand support across more AI platforms, account more accurately for uploaded documents and files, and provide increasingly model-specific environmental estimates.
We also want to make optimization more intelligent so PromptFootprint can learn when shortening a prompt is helpful and when additional context is actually necessary.
Ultimately, we envision PromptFootprint helping individuals, schools, developers, and organizations understand exactly how they use AI and how they can make that usage more efficient.
As AI adoption continues to grow, our goal is simple:
Fewer wasted tokens. Lower costs. Lower resource consumption. Better AI usage.
Log in or sign up for Devpost to join the conversation.