Search licensing blocked AI summaries

This reveals how tightly search distribution contracts can box in product strategy. Ecosia already licensed regular web search from Google and Microsoft, so building an AI answer layer on top could create a direct overlap with those vendors' own AI search products and expose Ecosia to contractual risk. That pushed the team toward an independent provider that could supply both retrieval and summarization, while moving fast with a six person team.

• The practical issue was not just getting links from a search API. Ecosia needed a service that could take a complex query, fetch web results, generate a summary with citations, and fit into a live consumer search workflow with acceptable latency. Exa was used on 30 to 40 percent of queries, about 500,000 times per day.
  1 sacra
• General purpose APIs would have split the stack in two. SerpAPI style products could return search results, but Ecosia would still need to build the summarization layer, routing logic, and serving infrastructure itself. For a small AI team trying to ship quickly, that meant slower time to market and more operational work.
  1 sacra 3 sacra
• This is also why vendor behavior mattered as much as raw search quality. Ecosia found Exa, Parallel, and others broadly similar on result quality, but chose the provider that offered custom pricing, engineering support, and product flexibility. In this market, compliance room and integration help can matter more than index quality alone.
  1 sacra 5 sacra

The next step for this market is a split between infrastructure suppliers that let search engines build differentiated AI layers, and incumbent search platforms that keep more of that layer for themselves. That gives independent providers an opening, especially with Bing Search APIs retired in August 2025, but the winning vendors will be the ones that pair search quality with flexible terms and close implementation support.