Traditional search makes you do the research. Wave does the research and shows the evidence.
- 01
Understand
Wave reads your question, and your earlier questions in the same session, so follow-ups like “only under $50,000” keep their context.
- 02
Search
Wave queries several independent web search backends at once, merges the results, removes duplicates and ranks them for relevance, authority and freshness.
- 03
Read
Wave fetches the top pages itself and extracts the passages that matter. If a page can’t be read, Wave marks the source “Snippet only”.
- 04
Answer
A GPT model writes a concise answer using only those pages. Every factual claim carries a numbered citation.
- 05
Show the evidence
Citations are built from the pages Wave actually retrieved. The model can point at a source, but it can never invent one, and Wave removes any citation that doesn’t match a real result.
What Wave won’t do
- Invent sources, links, quotes, experts or statistics.
- Claim to have read a page it didn’t retrieve.
- Quietly settle disagreements. When sources conflict, Wave says so and shows both sides.
- Present its own reasoning as a sourced fact. Inferences are labelled.
Answers can still be wrong. For health, legal, financial and safety questions, read the sources and check important decisions with a qualified professional.
Privacy
- No accounts, no ads, no tracking cookies.
- Wave doesn’t store the text of your searches in its analytics. It records only counts and timings, such as how long a search took and how many sources were cited.
- To answer, your question is sent to Wave’s search providers and to OpenAI’s API. Conversation context lives in your browser tab and disappears when you close it.
- Rate limiting uses a salted, one-way hash of your network address held briefly in memory, never written to disk.