Qwant's AI answers rely on web pages indexed in our search results. Here's how it works.
What sources does the AI rely on?
On the flash answer
When you perform an eligible search, the AI uses the results from our web index and produces a summary. The sources used are cited in the answer block, in the form of clickable links.
This approach (retrieval-augmented generation) ensures that the answer is based on existing and verifiable content. Not just on what a model memorized during its training.
On the AI Chat
The AI Chat also relies on Qwant’s web index to find relevant information and offer you a clear answer, with sources when available.
Unlike the flash answer, the AI Chat can also draw on the model’s general knowledge when relevant. For example, on more “cold” and stable topics (mathematics, definitions, general concepts), or for creative requests (ideas, reformulations, outlines…).
What data is transmitted to the AI?
Your text search is transmitted to the model that produces the answer. What is not transmitted:
- Your IP address is not sent to the model.
- No search history is used to personalize the answer.
- Your query is not associated with a user profile.
Furthermore, the query is not retained by the partner (Zero Data Retention); it is deleted after processing.
Are your data used to train the AI?
No. The searches you submit to Qwant are not used to train AI models. Our privacy commitment applies to all features.
Who provides the language model?
Qwant may rely on third-party models. The choice of partner is made in line with our privacy requirements and European hosting. Our partners are listed on about.qwant.com.
How much trust should be placed in AI answers?
Qwant’s AI is provided in an experimental framework and without any guarantee of accuracy. Answers may be incomplete, approximate, outdated, or sometimes obsolete, and do not reflect Qwant’s opinions. It is strongly recommended to verify an answer with the cited sources and other sources.