We think it's a fair question, and it deserves a straight answer rather than a defensive one. As a team that cares a lot about transparency (it's our tagline), we dug into what Ecosia has said, what critics have said, and what the numbers actually show.
What Ecosia actually launched
In December 2025, Ecosia introduced two AI features. Overviews are short, cited summaries that appear at the top of search results. AI Search is an interactive, chat-style mode. Both are optional, and users can switch AI off entirely and go back to traditional search.
Ecosia's position is that it keeps the footprint low by using small models and by steering clear of heavy features such as video generation. It also says it relies on tools like AI Energy Score and EcoLogits to pick efficient models and track their energy use.
The case that it's still consistent with the mission
Let's give the argument its due, because it isn't a weak one.
- It's opt-in, or at least opt-out. Nobody is forced into AI answers. Anyone who wants the classic experience can keep it.
- Ecosia runs on its own clean energy. The company says it generates more renewable energy than its AI features consume, and points to roughly €18M invested in renewable projects. Its 2024 report said it produced about twice the renewable energy needed to power all searches. Assets include a 13 MWp solar park near Rottenbach in Germany, built as a joint venture with Naturstrom AG.
- Per-query costs are shrinking. Google reports a median Gemini text prompt at about 0.24 Wh, and Sam Altman has cited roughly 0.34 Wh for a ChatGPT query. Epoch AI estimates a typical GPT-4o query at around 0.3 Wh, far below the older "3 Wh" figure that circulated for years.
- Users want it. Search is changing. A green search engine without AI risks losing people to less responsible alternatives, and with them, the revenue that funds tree planting.
The case that something doesn't add up
Here's where the discomfort comes from, and honestly, we share some of it.
- No published numbers. We found no published energy-per-query figure from Ecosia for its AI features or a total for its AI usage. Climate writer Ketan Joshi says he asked the company for both and was pointed to a blog post that contained neither. Other critics have called the messaging greenwashing.
- The clean-energy math is slippery. The "twice as much as we use" claim was calculated against search, and the solar investments pre-date the AI features. Saying you generate more than your AI uses is hard to verify when the AI side of the equation is undisclosed.
- Training and upstream costs are out of frame. Ecosia's models are built by others. The energy used to train them, and the emissions from manufacturing the hardware, mostly don't show up in a "per query" story.
- Cheaper queries can still mean bigger totals. This is the rebound effect. If AI answers become easy and everywhere, people ask more, and total consumption can rise even as each query gets more efficient.
- Offsetting isn't erasing. Clean generation on the grid is valuable, but it doesn't make added demand free.
Estimates here vary enormously. A short text prompt may cost very little, while a long, document-heavy one can cost many times more. Water estimates range from a fraction of a millilitre to hundreds of millilitres per prompt depending on methodology. Anyone quoting a single confident number, for or against, is simplifying.
The bigger picture
Zoom out and the stakes get clearer. The International Energy Agency's base case sees global data-centre electricity use rising from roughly 415 TWh in 2024 to about 945 TWh by 2030, with AI-focused servers growing far faster than conventional ones. One search engine's choices won't move that curve. But Ecosia isn't any search engine. Its entire brand is the claim that technology can be part of the climate solution, which means it gets held to a higher standard.
So, is it defeating the point?
Our take: not necessarily, but it hasn't earned the benefit of the doubt yet.
Adding AI does not erase Ecosia's tree-planting work. A small-model, opt-out approach backed by real renewable investment is a more defensible position than most of the industry can claim. But "trust us, it's green" isn't enough for a company whose whole identity is built on trust. Without published energy-per-query data and a clear accounting of AI's total footprint, users are being asked to take the most important part on faith.
The good news is that this is fixable. If Ecosia published its AI energy figures, named the models in use, and reported usage and generation side by side, it would turn a debate about vibes into one about evidence. That's the standard we'd want from any infrastructure we depend on, and it's the one we try to hold ourselves to at PulseWatch.
In the meantime, if the question matters to you, the practical step is easy: check your settings, decide whether you want AI in your search at all, and keep asking for the numbers.
Comments
Questions, corrections or thoughts on this post? Join the thread below.