Carbon Canvas
Free · private · no signup

See the energy range behind your AI use.

Build an individual, course or voluntary cohort scenario. Carbon Canvas shows a transparent range—not a falsely precise footprint.

Runs in your browser No prompt or reply content Versioned assumptions
01 · Scenario

Describe typical use

Local only
Who is this estimate for?Choose a scope
Start with a patternYou can fine-tune it
Estimated monthly AI energy
≈ 69–823 Wh

About 30 phone charges, spread across a month.

Low–medium confidence Method e-0.2.0

Based on about 433 messages a month (typical message).

Water perspectiveroughest estimate
≈ 0.0–5.8 litres

Data-centre cooling and electricity-generation water vary enormously. This is shown only for perspective and is not included in your footprint.

Why is this a range?
01The exact model, hardware and batching behind each message are usually unknown.
02Data-centre location, power mix and cooling system are not disclosed per request.
03Message length and hidden reasoning can change energy use substantially.
→Carbon Canvas compounds those uncertainties instead of presenting a made-up exact number. Inspect the methodology.

Saved scenarios

Save a scenario to compare it later. History stays in this browser.

This is a scenario. Want automatic tracking?

The Carbon Canvas extension is being prepared to estimate AI activity privately in your browser without reading conversations.

Join the extension waitlist

These are rough scenarios for learning and pilot planning—not observed usage, a formal carbon audit or reporting-grade emissions data. Cohort mode multiplies a representative per-person pattern and does not prove institutional AI use. Energy is the strongest estimate; water is substantially rougher. See the published methodology and limitations.