Artificial Intelligence (AI) is everywhere. From the chatbotI have no idea what this means helping you with your homework to the recommendation algorithm deciding which video you watch next on TikTok, it has become a huge part of our digital lives. But while we marvel at how smart these systems are, we rarely stop to ask: what is this actually costing the planet?

To understand the environmental impact, we first need to look at how AI works. It isn't magic; it is massive amounts of computing power. This happens in two distinct stages. First, there is Training. Before an AI can answer a single question, it must be "taught" by feeding it billions of sentences or images. This requires supercomputers running 24/7 for months, consuming huge amounts of electricity and water for cooling. Once trained, the AI enters the Inference stage. This is the part you use. Every time you type a prompt into ChatGPT or ask Gemini to write a poem, the AI has to "think" (calculate) to generate an answer. While training costs a lot upfront, the cumulative effect of millions of people using AI every day for inference is where the real long-term energy cost lies.

The Energy Cost: Crunching the Numbers

It is easy to think that digital actions are "free" because we can't see the exhaust fumes. But every tap, click, and prompt sends a signal to a data centre that burns electricity.

When we compare AI to a standard Google search, the difference is stark. A standard web search uses roughly 0.0003 kWh of energy. In contrast, a text-based AI query (like ChatGPT) uses roughly 0.003 kWh to 0.009 kWh. This means asking an AI is roughly 10 to 30 times more energy-intensive than just Googling the answer yourself.

The "Forced" AI Issue and Zombie Energy

This comparison highlights a new controversy in the tech world. You might have noticed that search engines are beginning to generate AI summaries automatically at the top of your search results, whether you asked for them or not.

This creates a problem known as Zombie EnergyI have no idea what this means. If you search for a specific website and click the top link immediately, you likely ignored the AI summary box. However, the data centre still burned the electricity - spending 10 times the energy of a normal search—to generate that summary before you clicked away. With search engines handling billions of queries a day, this "forced" AI usage creates massive amounts of wasted energy on content that no one actually reads.

Text vs. Images vs. Video

It is also important to remember that not all AI is created equal. Generating text is relatively "cheap" in energy terms, roughly equivalent to charging a smartphone for a few minutes. Image Generation is significantly heavier; creating just one AI image uses about as much energy as fully charging a smartphone. Video Generation is the most demanding of all, requiring incredible computational power and carrying a much higher carbon footprint.

The "Real World" Comparison

Before we decide to ban all robots, we need some perspective. How does using AI compare to other things we do every day without thinking?

Boiling the Kettle
We love tea in the UK. Boiling an average electric kettle uses about 0.1 kWh of energy. To put that in digital terms, boiling the kettle just once uses the same amount of energy as roughly 20 to 50 AI text queries or 300 standard Google searches.

Streaming Movies and Social Media
Streaming video is one of the biggest data hogs on the internet. Watching 1 hour of Netflix (HD) uses roughly 0.08 kWh to 0.2 kWh. This means watching a movie is roughly equivalent to boiling the kettle twice or asking an AI 50+ questions. Similarly, scrolling through image-heavy or video-heavy feeds like TikTok or Instagram for an hour consumes significantly more energy than a quick 5-minute chat with an AI to help explain a coding concept.

So, Should I Be Worried?

The short answer is: No, but you should be conscious.

If you panic about using AI for your homework but then spend 4 hours streaming 4K video while leaving the lights on downstairs, you are focusing on the wrong thing. However, the "rebound effect" is real. Because AI is so easy to use, we might start using it for everything, dramatically increasing our total energy usage. If 100 million students ask AI "What is 2+2?" instead of doing it in their heads, that is a lot of wasted energy.

How to be a Green AI User

We can still use these amazing tools, but we should use them efficiently.

First, vote with your clicks. If a search engine forces an AI summary on you that you don't need, scroll past it. User metrics teach companies which features are actually useful versus which ones are wasting resources.

Second, don't use a sledgehammer to crack a nut. Do not use a massive AI model (like GPT-4) for simple factual questions like "What is the capital of France?" Use a standard search engine or, preferably, your own knowledge for that.

Finally, you should master Prompt Engineering. If you have to ask the AI 10 times because your first prompt was vague, you have used 10 times the energy. Learning to write clear, specific prompts to get the right answer the first time is the single best way to reduce your AI carbon footprint.

Conclusion

AI is a powerful tool for learning. Like electricity or running water, the goal isn't to stop using it, but to avoid wasting it. By being a smart user—asking better questions and knowing when not to use AI—you can get all the benefits without the guilt.