I Use AI Every Day.
Is That Actually the Worst Thing for the Planet?

A data-driven investigation into one household's annual carbon footprint — and the finding that completely reframes where the guilt should go.

There's a particular kind of modern guilt that comes with being a heavy AI user. Every prompt feels like it's spinning up a server farm somewhere, guzzling electricity, melting glaciers. I use Claude and ChatGPT constantly — for work, for coding side projects, for planning, for writing. On some days, AI is as essential to my workflow as coffee.

So I decided to do what any data-curious person would do: actually run the numbers. What I found surprised me — not because my footprint was shockingly high or impressively low, but because of where the emissions actually come from. The answer reshapes the entire guilt conversation.

This is not a guilt piece. This is an investigation.

Who We're Investigating

A household of five in Denmark — two adults, three children. Both adults work in Aalborg, with a combined daily commute of approx. 25 km each way. One works in research administration at the university; the other works nearby. The household runs on one electric car. Family roots in the Philippines and Thailand. Financially conscious. Moderate consumption habits. Heavy users of digital tools. Not trying to be saints. Not ignoring the problem either.

The Numbers at a Glance

Three scenarios — based on the family's realistic flight pattern: no Asia trip this year, a European holiday, and a full-family long-haul trip to Asia (which happens roughly every two to three years).

The Danish average is approximately 11–12 tCO₂e per person per year (consumption-based). EU average: ~8–9 tCO₂e. Global average: ~4.7 tCO₂e.

How Does That Compare?

That last bar is not a rendering error. It is genuinely that small.

Breaking It Down, Category by Category

Home Energy

Denmark's electricity grid runs on roughly 55–65% wind power, which meaningfully lowers the emissions from household electricity. District heating — common in the Aalborg area — is significantly cleaner than gas.

Transport

The household runs on one electric car, shared across two working adults. The daily commute covers approx. 25 km each way — a distance that, on Denmark's clean grid, produces a fraction of what the same trip would emit in a petrol vehicle.

For comparison: the same commute in an average petrol car would produce approximately 0.99 tCO₂e annually — almost three times higher. The EV is already one of the better decisions in this footprint.

Aviation — The Variable That Changes Everything

This household flies. Not recklessly, but regularly — a European trip most years, and every two to three years, all five make the long haul to Asia. That rhythm has real carbon consequences, and the numbers deserve to be read clearly.

Emission factors include the Radiative Forcing Index (RFI ×2) — a standard multiplier applied to aviation because high-altitude NOx emissions, contrails, and cirrus cloud formation roughly double the effective warming impact beyond CO₂ alone. This is the DEFRA and IPCC recommended approach.

Food

An active, mixed diet with intermittent fasting. Not fully plant-based, but not a heavy red-meat household. Three children's food is included. The biggest lever here — as always — is beef and lamb frequency.

Shopping & Consumption

This household actively prefers digital products over physical ones, builds software tools instead of buying commercial equivalents, and makes purchases deliberately rather than habitually. Three children add some baseline consumption that can't be avoided.

Digital Life & AI — The Category You Were Probably Expecting to Be Worse

Here is what an AI prompt actually costs. A single query to Claude or ChatGPT uses approximately 0.001–0.01 kWh of electricity, depending on query length and data centre efficiency. At a mid-range estimate of 0.003 kWh per prompt, and assuming large cloud data centres operate on a grid mix of roughly 200–400g CO₂/kWh:

The Full Picture

What Actually Moves the Needle

The Highest-Impact Changes, Ranked Honestly

1. Asia trip frequency — by an enormous margin. Every additional trip to Thailand or the Philippines adds approximately 17.5 tCO₂e for a family of five. Extending the cycle from every two years to every three reduces annualised aviation emissions by roughly a third. No other single decision comes close.

2. European holiday routing. A family of five flying short-haul annually adds ~1.75 tCO₂e. Train alternatives to nearby European destinations — overnight to Paris, rail to Germany or the Netherlands — can eliminate most of this for many trips. Not always practical, but worth knowing.

3. Reduce beef and lamb specifically. Poultry, fish, and plant-based proteins carry 5–20× lower emissions per kilogram than beef. Partial substitution across a family of five makes a meaningful compound difference over years.

4. Keep the EV, and keep driving it. The electric car is already one of the better structural decisions in this footprint. On Denmark's grid, it performs roughly three times cleaner than a petrol equivalent for the same distance. No action needed — just continuation.

5. Ignore the AI guilt entirely. It is not measurable at this scale relative to anything else on this list.

Is Building Software a Greener Hobby?

This household's side projects — a family task board, a grant management tool, a training planner, a personal blog — are all digital. No materials, no shipping, minimal physical goods. Compared to common alternatives:

Building digital tools is among the lowest-carbon hobby categories that exist. The comparison is not even close. The marginal electricity cost of running a small web app, a PWA, or a personal automation is genuinely trivial — and the activity replaces consumption that would otherwise have a physical footprint.

Five Things People Get Wrong About AI and Carbon