NeevaAI Takes Its Cited Search Answers to More Countries

Neeva expanded NeevaAI to logged-in users in Canada, the U.K., Germany, France and Spain, while its regular search engine was set to reach Australia and New Zealand. The service combines answers from web sources with citations and falls back to links when it is not confident enough to answer.

WTF Index NEUTRAL
◄ Terminator 1 Idiocracy 1 ►

NeevaAI expands a routine search product with citations and cautious fallback behavior, with only a mild concern about answer quality.

NeevaAI Takes Its Cited Search Answers to More Countries

Neeva is taking its AI-powered search experience beyond the United States as competition grows around search engines that generate answers instead of displaying only a list of links. The expansion brings NeevaAI to logged-in users in Canada, the U.K., Germany, France and Spain, while Neeva’s regular search engine is also rolling out to Australia and New Zealand.

Answers with sources attached

Traditional search results ask people to scan links and assemble an answer themselves. NeevaAI is designed to do more of that work: it searches for information across multiple sources, then generates a response with citations pointing back to those sources.

Those citations are part of the product’s approach to transparency. They let users see where an answer came from, and can give the publishers of the cited material credit and a chance to receive referral traffic. When NeevaAI cannot confidently answer a question, it can instead show website links and excerpts for the user to investigate.

In tests described by TechCrunch, the tool handled questions about why the Beatles were popular and the world 5k running record by drawing on multiple cited sources in real time. The results were not presented as proof that the system could answer every kind of question.

A cautious approach to uncertain questions

Neeva co-founder and CEO Sridhar Ramaswamy said the system is intended to answer questions for which it can find authoritative websites. The company’s stated priority was avoiding unsupported or incorrect answers, so the search experience may decline to respond when it lacks confidence.

That can mean a question containing a mistaken premise gets no answer instead of a correction. The article gives the example of asking when Boris Johnson served as the king of the United Kingdom. A more helpful response might explain that the question should refer to his time as prime minister, but NeevaAI’s cautious behavior could leave the user without that clarification.

Ramaswamy said the company planned to improve how the system handles such cases. At the time of the report, however, the details of a future response were not clear. The tradeoff is visible: refusing to guess can limit false claims, while a refusal may also miss an opportunity to guide someone toward a better question.

Search competition shifts toward generative AI

The international rollout came as major technology companies were also introducing AI features for search. Microsoft had recently announced a new Bing experience powered by a more advanced version of ChatGPT and customized for real-time search. Google had unveiled Bard, an experimental conversational AI service intended to provide synthesized responses gathered from multiple sources.

These products point toward a different search interaction: people ask a question, the service looks across available material and presents a composed response, often alongside links or citations. That may save users from opening several pages, but it also makes the quality and traceability of the answer central to the experience.

Neeva had launched NeevaAI in the U.S. in January, following the introduction of its search engine in the U.S. in June 2021. The company had also brought its search engine to Europe and added a free basic plan alongside its subscription offering.

Privacy and independence are part of Neeva’s pitch

Neeva’s business model is another part of its attempt to stand apart. The company says it does not monetize through advertising and blocks third-party trackers from using personal data to show personalized ads. It aims to earn revenue through subscriptions, while its basic tier includes restrictions.

The company was also building its own search technology, though it still relied on Bing for some web search activity. For NeevaAI, Ramaswamy said the company worked with large language model providers including OpenAI’s GPT-3.5 and Claude from Anthropic. He said Neeva used those systems for some training data and summaries, while most models used for NeevaAI were its own, trained or fine-tuned for tasks such as question-answering and summarization.

Neeva therefore entered a crowded field with a combination of generated answers, citations, an ad-free business model and an effort to build its own search stack. The expansion put that approach in front of more users, while leaving the company to show whether cautious answers and privacy-focused search could distinguish it from larger competitors and other startups.