A Simple Comparison Using Everyday Classroom Activities

When you search the internet or ask a question to an AI system, energy is used to process your request. Some actions use very little energy, while others require far more. To help make sense of this, we compare AI queries to something everyone understands: photocopying.

The Short Answer

• 1 standard Google search uses a very small amount of energy.
• 1 AI query (ChatGPT, Gemini, etc.) uses around 10 times more energy than a normal search.
• Photocopying, once you include paper manufacturing, toner, and electricity, uses more energy than both digital tasks.

This article explains how they compare.

1. Energy Used per Query

We measure energy in Watt-hours (Wh).

Standard Search

0.3 Wh per search
This is extremely efficient because search engines use indexes to retrieve information quickly.

AI Query

3 Wh per query
A large language model generates an answer word by word, requiring billions of calculations.

2. The Real Energy Behind a Photocopy

Photocopying is often assumed to be energy-light, because the machine itself uses little electricity.

However, a proper carbon and energy footprint includes:

• the electricity used by the photocopier,
• the energy used to produce the A4 paper,
• the energy and materials in toner manufacturing,
• packaging and transportation.

Environmental studies give a typical A4 photocopy a total footprint of:

• 5 to 20 Wh per sheet of energy equivalent
• 4 to 6 grams of CO₂e per sheet

To keep comparisons simple, we use a reliable mid-range value:

One A4 photocopy ≈ 15 Wh (total lifecycle impact)

This means photocopying is mainly an environmental issue not because of electricity, but because of paper production.

3. AI Queries vs Photocopying

Using the figures above:

• 1 AI query (3 Wh)
  ≈ 0.2 photocopies
• 1 photocopy (15 Wh)
  = about 5 AI queries

Put another way

4. Other Everyday Equivalents

These comparisons help students and teachers understand digital energy use in context.

Charging a Smartphone

8 to 12 Wh
≈ 3–4 AI queries

Running a 3 W LED bulb for 1 hour

3 Wh
≈ 1 AI query

Boiling a kettle for 1 minute

33 Wh
≈ 11 AI queries

Streaming HD video for 1 hour

50 to 100 Wh
≈ 17–33 AI queries

Final Summary

• A standard search uses very little energy.
• An AI query uses about 3 Wh, similar to running a small LED bulb for an hour or one-third of a phone charge.
• A photocopy, when you include the impact of paper and toner, uses five times more energy than an AI query.
• This means 100 AI queries have roughly the same total energy impact as 20 photocopies.

References

De Vries, A. (2023). The growing energy footprint of artificial intelligence. Joule, 7(10), 2191–2194. https://doi.org/10.1016/j.joule.2023.09.004

Dias, A. C., & Arroja, L. (2011). Comparison of methodologies for estimating the carbon footprint – case study of office paper. Journal of Cleaner Production, 24, 30–35. https://doi.org/10.1016/j.jclepro.2011.11.005

Electricity 2024. (n.d.). World Energy Outlook 2025. Retrieved November 19, 2025, from https://www.iea.org/reports/electricity-2024

iPhone battery and performance – Apple Support (MY). (2025, April 16). Apple Support. https://support.apple.com/en-my/101575

Power usage effectiveness – Google Data Centers. (n.d.). Google Data Centers. https://datacenters.google/efficiency/

Signify Holding. (2023). State-of-the-art LED light bulb for the home.