A ‘Green’ Search Engine Sees Danger—and Opportunity—in the Generative AI Revolution

In the era of search wars fought between giants, it’s tough to be small. Berlin-based Ecosia offers a search engine for the climate-conscious, promising to be carbon-negative by investing all of its profits into planting trees—more than 180 million of them since it launched in 2009. It’s not likely to topple Google, but it has won a stable clientele of around 20 million users with that green branding and by repackaging search results from Microsoft’s Bing. But after a decade of little change in the search business, everything is now in flux, thanks to generative AI. “I’ve never seen so much change in the market as in the last six months,” says Christian Kroll, Ecosia’s CEO.

The tumult has forced Ecosia to rethink its business plan in order to compete with new chatbot-like search engines built on large language models. Today, the company began switching from providing results exclusively from Microsoft’s Bing, as it has for the past 14 years, to primarily sourcing them from Google—though it will still syndicate some Bing results via marketing company System 1. At the beginning of the year, Kroll says, Ecosia “got some signals from Microsoft that kind of triggered us to be a bit more on the lookout for other potential providers.” In March, Microsoft hiked its prices for search results, which was “a wake-up call for alternative search engines,” according to Kroll. Microsoft declined to comment.

Ecosia switched partners in hopes of finding a way to participate in the profound shift in how people search the internet triggered by AI. The company is only testing its partnership with Google and isn’t immediately going to be using the search giant’s AI tools—though it hopes to do so in future.

For a small provider like Ecosia, the recent disruption in search could be an opportunity to reach new markets and offer new services to users and advertisers. But the shifting landscape is also fraught with challenges. Although there are startups working on AI-powered search, the category is still mostly a competition between giants. AI-generated search results also create new legal and ethical issues for providers to solve. And for a search engine that gives away all of its profits to fight climate change, there’s the problem of a step-change in energy use needed to power generative-AI.

“This complexity means we have a lot more topics to deal with now,” Kroll says. “As a small company we have to place our bets carefully. Google and Microsoft have a lot more coins that they can spend in the casino.”

Microsoft, which has invested a reported $13 billion into ChatGPT developer OpenAI, launched a chatbot-style interface for Bing in February. A month later, Google launched its Bard chatbot in the US and UK. Conversational generative AI like ChatGPT changes the way a user interacts with search, and the way that results are presented. The last generation of search engines responded to a user’s query with a list of links to other media where they could find a detailed answer. AI-powered search attempts to answer the question for itself.

“The trend that we're already seeing is that the idea of having just 10 blue links on a search result page isn't really going to cut it anymore,” Kroll says. “If you look at the numbers, less and less traffic is actually leaving the search results page.” Increasingly, people are transacting directly via search results on the big search engines too, Kroll says, such as by making hotel or flight reservations without leaving the portal.

If Ecosia can catch hold of that trend of search engines being more than just an intermediary, it could make more revenue by taking a cut of users’ transactions. Being more actively involved in people’s online transactions could also provide an opportunity to nudge them into making more environmentally conscious decisions.

Kroll says one place Ecosia could intervene to suggest greener choices is on searches for flights. “We could tell them what the cheapest flight is, but also that they could take the train instead and how much CO₂ they could save,” he says. The era of generative search could offer new ways for a specialist search engine to stand out. “If you’re giving only one answer, then it’s even more important that you have a moral stance on that answer as well.”

While AI may offer search providers new opportunities, it also takes them into complicated legal territory. In the US, search engines are protected from legal liability for any harm caused by the results that they surface, under Section 230 of the Communications Decency Act. However, it’s not yet clear whether that will apply to an answer given by a chatbot-powered search engine. New AI regulation is coming in Europe, which may impose restrictions on how AI-generated content is labeled and presented to users.

This is territory that Kroll admits he isn’t entirely sure how to navigate. The company might need to disclose its biases, he says, and openly say that it's trying to encourage users to make decisions with better environmental outcomes. But that’s complicated by the fact that the AI that generates those results—which Ecosia would have to license from a larger tech company—would be a black box to him and his colleagues. When a smaller company licenses search or AI results, how they are generated and what has been filtered out isn’t generally provided. “We don’t get that information,” Kroll says. “That’s a problem.”

Another problem that is particularly acute for Ecosia, which displays a counter on its homepage telling users how many trees it has planted, is the environmental footprint of generative AI. By some estimates, AI-enhanced search results require a fivefold jump in computing power compared to conventional search technology, because of the extra work required to train and deploy generative AI models. That, almost inevitably, means a big rise in power consumption and, potentially, in carbon emissions. Many data centers use renewable energy, and Google aims to run its facilities on carbon-free energy 24/7 by 2030, but manufacturing the powerful chips and other components needed to power generative AI projects generates emissions too.

Kroll says it’s possible that generative AI might help offset some of its own environmental impacts by aiding projects that reduce societies’ emissions overall. And, since all Ecosia’s profits go into emissions reductions, he says adopting the technology still makes sense. At the moment, the company estimates that each search by a user generates 0.2 grams of CO₂, while the tree planting projects it funds sequester many times that. But the growth in energy consumption is something the industry as a whole is going to have to reckon with. “I mean, if we had all big tech companies running on the same principle [as us], we would have solved the climate crisis,” Kroll says.