You Use AI Daily — But At What Cost? 5 Steps For Smart AI Use

---

You are committed to being sustainable and making choices that are good for the environment — you recycle, perhaps even compost. You are mindful of your energy use; maybe you have installed solar panels or even switched to an electric car. But there is one habit quietly undermining all your good intentions. One you rely on daily, one you increasingly cannot do without. And that is your AI use — and how sustainably you use it.

Every AI query you run has an energy cost. As someone who teaches and writes about both sustainability and AI, I feel a tug between the two every single time I use it. The cognitive dissonance of championing a technology while worrying about its environmental footprint is real — and I suspect I am not alone.

The evidence is real and sobering: A typical day of moderate AI use — some text questions, a few image attempts, maybe a short video — can consume enough energy to run your microwave for three and a half hours. And if you reach for a powerful reasoning model rather than a basic one, that footprint can be 30 times higher — and in some cases up to 100 times higher — for the same task. Multiply that across 2.5 billion AI queries served every single day , and the scale becomes clear.

Yet despite growing awareness, most of us feel stuck. A 2025 survey found that 72% of Americans are concerned about AI's environmental impact — more than they are about air travel or meat production. Over 70% admit feeling guilty about their digital habits but carry on unchanged .

You don’t have to choose between being productive and being responsible. Here are five evidence-backed steps you can make to your daily AI use — starting today.

Step 1: Think Before You Type

It's Monday morning. You open your AI tool to help draft a quick email to your team. You type something like "write me an email about the meeting." The response comes back too formal, too long, or missing the point. You type again. And again. Fifteen minutes and six exchanges later, you may have something usable. What you may not have realized is that each iteration costs energy.

The fix is simple: be specific from the start. Try this instead — "write a short, informal email to my team of five summarizing the key decisions from this morning's budget meeting and the three next steps we agreed on." One precise prompt that includes the context, the format you want, and the audience it's for. You will get a better answer — and use less energy getting there.

The energy savings are real. Research from University College London (UCL) found that cutting prompt and response length by half can reduce energy consumption by up to 75% . A skilled prompt writer achieves four to five times better efficiency than someone who relies on trial and error — better results while using a fraction of the energy.

Step 2: Match the AI Tool to the Task

You need to summarize a two-page meeting report before your 3pm call. You open your go-to AI model — which also happens to be the most powerful — why settle for anything else — paste in the document, and ask for a summary. It works great, of course, as you knew it would. But what you may not realize is that you just used the equivalent of a commercial jet engine to power a bicycle ride around the block.

Does every task you use AI for — summaries, simple drafts, translations, quick questions — need the most powerful AI model available? Most likely no. Most major platforms today offer lighter, more energy-efficient options alongside their flagship models. On Claude, it’s called Haiku. On ChatGPT, it is GPT-4o mini. On Gemini, you will find Flash or Flash-Lite. These models are specifically designed for routine tasks and deliver comparable results at a fraction of the energy cost. The UCL study found that using smaller, task-specific models can cut energy use by up to 90% with no meaningful loss in performance.

Before you open your AI tool, ask yourself — does this task need the most capable model, or will a lighter version do the job just as well? Look for terms like "mini," "flash," or "lite" in your model menu — platform names change regularly, but the principle does not.

Step 3: Your AI Image Habit Is Costing More Than You Think

A picture is worth a thousand words. Our brains process images 60,000 times faster than text, and we retain 65% of what we see compared to just 10-20% of what we read. No wonder AI image generation has become one of the most popular uses of the technology. But here is the environmental reality: it is also by far the most energy-intensive thing we do with AI — and the gap between generating text and generating an image is far larger than you might think.

A standard AI text response uses around 0.24 watt-hours of energy — roughly the same as running your microwave for one second. Generating a single AI image uses five times as much . Ask AI to produce a short five-second video and the energy cost can be up to 4,000 times higher than that text response. Read that last number again.

None of this means you should never use AI to generate visuals. Sometimes it genuinely adds value — a custom illustration, a unique graphic that would take hours to create manually. The question worth asking before you click generate is simply this: does this visual actually need to be AI-generated, or is there a simpler option that works just as well? A stock image, a basic PowerPoint shape, a simple chart — these alternatives cost a fraction of the energy and often serve the purpose equally well.

Step 4: Do More With AI By Asking Less Often

You have errands to run. Would you drive to the grocery store, come home, then drive back out to the pharmacy, come home again, and make a third trip to the dry cleaner? Of course not. You would bundle them into one trip — it saves time, fuel, and effort. Yet when it comes to AI, most of us do the digital equivalent of separate errands all day long — one quick question in the morning, a single task at lunch, another request mid-afternoon, and a few more scattered throughout the day. Each one triggers a fresh exchange with an AI model. Each one carries an energy cost.

Batching your AI queries — grouping multiple tasks into one focused session rather than dipping in and out throughout the day — is one of the simplest and most effective ways to reduce your AI footprint. And it has an added bonus: it tends to produce better, more consistent results because you are giving AI fuller context about what you need all at once.

Research from Google Cloud confirms that grouping multiple requests together means AI delivers the same useful output with less energy — the same way a full dishwasher uses less water than running it half empty.

Try this. At the start of each day or a focused work session, jot down everything you need from AI. Then tackle it all in one or two focused sessions. A single prompt may look like this: "Draft a follow-up email to my client, summarize the attached two-page report in three bullet points, and give me five talking points for my 3pm presentation on Q3 results." One prompt. Three tasks. A fraction of the energy of three separate exchanges. Your to-do list gets done — and your footprint shrinks.

Step 5: Turn Off What You Are Not Using

AI is no longer the tool you have to reach for when you need it. It is quietly embedded in your email client, your browser, your calendar, your writing software, your project management tool, and your smartphone keyboard. Many of these features are on by default and most of us have never turned a single one off.

OpenAI's CEO Sam Altman has noted that even saying "thank you" to your AI tool comes with a carbon footprint. Multiply that across every background feature running silently on your devices right now — and the numbers add up faster than you think.

This matters because always-on AI features do not wait for you to ask them something. They run in the background, processing context, anticipating your next move, and consuming energy continuously — whether you notice them or not. Longer context windows, persistent memory features, and background AI assistants all add to your footprint silently, every hour of every working day.

Research confirms that limiting context windows and disabling unnecessary background AI features significantly reduces energy overhead — the computational work happening behind the scenes without your knowledge or consent.

The fix is surprisingly simple and takes less than ten minutes. Go through the settings of your most-used tools, from your browser settings to your productivity apps such as Microsoft 365 Copilot and Google Workspace. Turn off anything you do not actively and regularly use. Keep what genuinely saves you time. Think of it as a digital declutter — and one that happens to be good for the planet.

The Bottom Line

The solution is not to stop using AI. It is to use it with the same intentionality you bring to every other sustainable choice in your life. Recycle. Compost. Drive less. And now — add intentional AI use to that list.

Five small steps. Millions of professionals making the same choices. That is how we use the tools we love — without hurting the planet we need.

By Anjali Chaudhry, Contributor