"If I run prompts all day, what am I doing to the planet?" is one of the most common questions we get in AI workshops - and most of the answers floating around are built on outdated, mis-scoped, or simply invented numbers. So we built a tool that lets you work it out for yourself and check the headline claims against the best current evidence.
The short version: for an individual, AI use is environmentally trivial - a year of heavy daily prompting is a few kettle boils' worth of electricity. The real conversation is the aggregate, industry-wide data-centre build-out, not the prompts you skip.
What the Energy Explorer is
The Energy Explorer is a self-contained, interactive page you can open on its own or share directly. It pulls the AI-and-environment debate apart into the few questions people actually argue about, and grounds each one in sourced, stress-tested figures rather than viral statistics.
Open the Energy Explorer
Move the slider, pick a scenario, and see the energy, water, and carbon cost of a year of your own AI use - in units you can relate to.
Launch the Energy Explorer →What's inside
An interactive personal-use calculator
Set how many prompts you run on a typical working day and the calculator estimates your energy, water, and CO₂ over a year - then translates it into everyday equivalents like kettle boils, hours of streaming, smartphone charges, and minutes of driving.
Three myths worth clearing up first
The claims that surface in nearly every workshop, and why each is misleading as commonly stated: the "AI uses 10× a Google search" figure (built on an old model versus a 2009 search number), the precise-sounding "Copilot uses 0.31 Wh per query" (a figure Microsoft never published), and the "every chat drinks a 500 mL bottle of water" claim (a 2023 GPT-3 estimate that conflates on-site and power-station water).
A traffic-light assessment of twelve claims
The twelve most-repeated claims about AI and the environment, each rated green, amber, or red on how well the evidence holds up - click any claim to see what's actually being claimed, the best current evidence, and what's being done about it.
Method, confidence, and sources
How the ratings were made - convergence, provenance, and arithmetic checks - and an honest account of where the uncertainty genuinely lies, with key sources from Google Cloud, Epoch AI, the IEA, LBNL, and peer-reviewed research.
Using it in a room? The Explorer was designed as a workshop reference - open it live, drop in the audience's own prompt volume, and let the numbers do the reassuring. Get in touch if you'd like us to run an AI session for your team.
The Energy Explorer was researched, built, and is maintained in-house by FutureAbility - part of how we help organisations separate AI signal from AI noise.
Figures reflect the best public evidence at time of writing (June 2026) and carry genuine uncertainty - treat them as well-grounded orders of magnitude, not precise constants. Full sources are listed within the Energy Explorer.