Multi-Source Search
How AI Search aggregates Wikipedia, news, finance, Brave, Diffbot and Exa into one progressive, streamed answer.
How aggregation works
When you run a query, AI Search fans it out to multiple sources in parallel rather than relying on a single index. Results are deduplicated, ranked and handed to the answer model with citations, so the response you read is grounded in what the sources actually returned.
| Source | Best for | Typical content |
|---|---|---|
| Wikipedia | Background and definitions | Encyclopedic summaries and entity facts |
| News APIs | Current events | Recent articles and headlines |
| Finance APIs (Alpha Vantage) | Markets | Stock prices and financial data |
| Brave Search | General web coverage | Web pages and fresh results |
| Diffbot | Structured extraction | Clean article and entity data from web pages |
| Exa | Semantic web search | Conceptually relevant pages, not just keyword matches |
Progressive answers
You never wait for the slowest source. AI Search returns a fast initial answer from the quickest sources, then enhances it as deeper results arrive — additional citations, refined figures and expanded context appear in place without re-running the query.
- 1
Initial answer
A fast first response streams in token by token, built from the sources that respond first.
- 2
Enhanced results
Slower, deeper sources finish and the answer is upgraded with more detail and citations.
- 3
Follow up
Ask a refinement in the same session — context is retained, so you do not repeat yourself.
Streaming responses
Answers are generated by leading LLMs and streamed as they are produced, so you start reading immediately instead of staring at a spinner. If the early part of the answer already covers what you need, you can stop generation and move on.
The answer model is given the retrieved source material for your query. Statements are tied to citations you can open and verify — if you want a stronger guarantee, run the claim through Fact-Checking.
Rate limits by plan
Search throughput is tiered by subscription. Free covers casual use; Basic, Pro and Enterprise raise the ceiling for heavier research workloads. When you hit a limit, AI Search tells you exactly when the window resets.
If your team regularly exhausts the Pro tier, Enterprise offers the highest limits plus priority access to all search sources. Compare tiers under Account → Subscription.