Should Animal Advocates Feel Guilty About Using AI?
The honest answer is more complicated and more hopeful than you think.
Maya is a fictional character, but her situation is one we hear often.
Maya runs a farm animal sanctuary. She's also a passionate climate advocate. Lately, she's been using AI tools to draft grant proposals, streamline donor outreach strategy, and research farmed animal welfare legislation. It saved her hours every week.
But then she read about data centers. The communities dealing with strained water supplies. The retired coal plants being brought back online to meet new energy demands. The local residents at town hall meetings, worried about what was coming to their neighborhoods. She recognized that look: the worried faces of people living next to something powerful that didn't care about them. It's the same look she sees on the faces of people who live near factory farms.
She hasn't used her AI tools since.
In the weeks that followed, she missed a grant application deadline. A piece of legislation affecting egg-laying hens moved through committee without her organization's testimony. Her weekly donor newsletter — the one that takes half an hour with AI and two without — didn't go out twice.
The environment didn't benefit at all.
If you're Maya, or if you recognize yourself in her story, this article is for you.
Two Truths You Need to Hold at the Same Time
Truth #1: Data centers do cause real, measurable harm to specific communities.
They can strain local water supplies, especially in dry regions
They concentrate enormous energy demand in places not built for it
The people most affected are often already vulnerable
Truth #2: For animal advocates, using AI does far more good than harm.
Every hour AI saves you, is an hour you can redirect to reduce animal exploitation, which is one of the most powerful environmental levers available to us
The environmental benefit of animal advocates who reach further and change more minds compounds in ways that far outweigh the seconds of electricity it cost to get there: fewer animals exploited, fewer factory farms that devastate local communities, less land destroyed, less water consumed
That tension doesn't mean one truth cancels the other out. It means the issue is real and deserves a real response, not a simple one. Let’s look at what the numbers actually show.
Energy
AI data centers use electricity to power GPUs — the chips that process your prompts. Those chips generate heat, which requires cooling systems to manage. Both the computing and the cooling draw energy.
In August 2025, Google published the most detailed analysis yet of its Gemini AI model (source). The result: a text query uses just 0.24 watt-hours of electricity, about ten times lower than estimates from just a year before. OpenAI's Sam Altman reported a similar figure for ChatGPT: 0.34 watt-hours (source). Independent researchers at Epoch AI and MIT Technology Review have confirmed numbers in the same range (Epoc AI, MIT).
So, knowing that AI is getting dramatically more efficient, here's what 0.24 watt-hours looks like in your life (source):
One AI prompt = watching TV for nine seconds
100 AI prompts = 15 minutes of television
One year of regular AI use (around 8 prompts/day) = running a space heater for a total of 30 minutes
To match the carbon of a single transatlantic flight, you would need to send one AI prompt every two minutes for 80 years straight
One important caveat: AI-generated video is a different story (source). Estimates suggest generating a 5-second video can use the same amount of energy as 2.5 hours of watching TV, while this number is noticeably higher than for text-based AI use (drafting, research, analysis), it helps put it into perspective of our existing energy consumption habits..
Water
AI data centers also use water to cool the same GPUs. Water runs through the equipment to carry heat away, then is either evaporated or drained back into local supplies. This is worth understanding clearly, because the headlines rarely explain it.
First, most water that passes through a data center is evaporated back into the broader water system unpolluted, not consumed or contaminated. This is meaningfully different from how agriculture uses water, which involves significant pollution of local supplies. The concern about data center water use is real but most acute for facilities built near fragile or already-stressed local water sources which is a policy problem for allowing permits to build data centres in such environments.
Second, the widely-cited statistic that "one AI prompt uses a bottle of water" is almost certainly wrong (source). The actual figure is closer to 30 ml per prompt; and roughly 85% of even that small amount comes from water used in electricity generation generally, not from the data center itself. Every electronic device you use draws on that same water-in-energy-generation system. Reporting AI's water use without that context is misleading.
Here's how AI prompts compare to other online actions and everyday choices (in milliliters):
Sending an email = 10 ml
Asking ChatGPT a question = 30 ml
Downloading a phone app = 40 ml
One hour of Zoom = 1,720 ml
Ten minutes of 4K video = 2,580 ml
Eating one beef burger = equivalent to 200,000 AI prompts
Every time you choose a plant-based burger or help someone else choose one you save more water than you could save in a lifetime of avoiding AI tools.
That said, the community stories you've read are real and there's a structural reason for them worth naming. Data centers are more efficient per computation than almost any other way of doing large-scale computing. But because they're so efficient, companies build them bigger and bigger which creates a strain on the local grid, the local water supply, and the local community. This is sometimes called the Jevons Paradox, and it's a legitimate concern at the systemic level which is why policy matters more than personal boycotts.
There's a Third Truth Nobody Is Saying
Avoiding AI to protect the environment, while working in animal advocacy, may be one of the most counterproductive trade-offs in the nonprofit sector.
Animal advocacy organizations are among the most resource-stretched in the nonprofit world. Most operate with tiny teams, tight budgets, and a workload that could consume three times the staff they have. Every hour lost to a task AI could handle in minutes is an hour not spent building campaigns, cultivating donors, training volunteers, or influencing policy to increase your impact, reduce animal suffering, and end factory farming.
Real Life Example: Friends of the Earth used AI to cut school menu analysis from 30-60 minutes per school to 1 minute. That single efficiency gain expanded their reach from 5 districts in California to 50 across the US, 25 in Arizona, and 25 in Mexico, giving them the national and cross-border data needed to make stronger cases to districts and directly reduce institutional demand for animal products.
This reduction matters as animal agriculture is one of the most environmentally destructive industries on the planet:
It contributes an estimated 20% of global greenhouse gas emissions (source)
It is the leading driver of deforestation (source) and freshwater consumption (source)
A single successful campaign, such as a state prohibition on foie gras production or school districts ordering less animal products, directly affects millions of animals and meaningfully reduces the environmental footprint
An animal advocacy organization that uses AI effectively will win more campaigns, reach more people, and move more policy than one that doesn't. More effective animal advocacy means fewer factory farms, which means measurably less greenhouse gas, less water use, and less land destruction.
The energy cost of your AI tools, measured in seconds of microwave time, does not come close to offsetting that.
Where Your Energy Actually Matters
Your advocacy work is already one of the most environmentally impactful things you can do.
Every campaign your organization wins, every corporate commitment, every legislative advance, every person who reduces their meat consumption because of your work, has a measurable positive impact on the planet.
On using AI well inside your organization
Do use AI for text-based work. Drafting, research, summarizing, translating, donor communications, policy tracking. The energy cost is negligible and the time savings are significant.
Consider non-generative AI tools for video. This is where energy costs become more noticeable. Tools that make color corrections and add filters to your video, or create shorter clips from longer videos use significantly less energy than generating a new video even if they use AI (for example OpusClip).
Choose AI providers with credible climate commitments. This sends real market signals and aligns your tool choices with your values.
Search for providers who show evidence that they use data centers with renewable‑energy guarantees.
Check the AI Energy Score Leaderboard before selecting a provider / model.
Research AI providers or models that selectively activate sub-networks – an LLM design called Mixture of Experts (MoE), as these are significantly more energy efficient.
Consider locally-run AI models. Some AI applications run directly on your device rather than in data centers (for example Ollama). If you're using renewable energy at home or in your office, this can reduce your footprint further.
Use lighter models where they're sufficient. Not every task needs the most powerful model available. A UNESCO report found that model choice alone can reduce energy usage by up to 90% (source) – an additional benefit is that lighter models are also cheaper!.
Build a simple internal AI policy. Not to restrict AI use, but to make sure it's directed toward your highest-impact work and that your team understands where the environmental tradeoffs lie.
For those who want to directly address data center impact
Consider supporting policy changes that prevent exploitations of communities near data centers
Push for water use disclosure laws for data centers in water-stressed regions
Advocate for renewable energy requirements before data center permits are approved
Support community notification rights when large data centers are proposed near residential areas
The Bottom Line
The cruelest irony would be if animal advocates sit out the most productive technology of our movement to protect an environment that animal agriculture is actively destroying. Too much is at stake to leave your best tools unused.
In our story, Maya restarts using AI tools. She submits a grant she nearly missed and is able to expand her sanctuary to take in 3 more goats. She uses the hours she saved to draft testimony on an upcoming farm bill that threatened to weaken farmed animal protections. And she still has time to contact her local elected officials to advocate for data center water disclosure requirements.