Web Search in Chat

By default, AI models only know what was in their training data. With web search enabled, Hubrix queries the web in real time before the model answers, so you get up-to-date information with source links.

Enabling web search

Look for the search icon (magnifying glass) in the chat toolbar — just below the message input area. Click it to toggle web search on. The icon highlights when active.

Web search stays enabled for the remainder of that chat session. It resets to off when you start a new chat.

How it works

When you send a message with web search on:

1. Hubrix extracts a search query from your message.
2. It calls the Serper API to fetch the top search results.
3. The page summaries and URLs are injected into the model's context.
4. The model answers using both its training knowledge and the fresh search results.

Citations appear inline in the response as numbered references — for example, [1], [2] — with full URLs listed at the end of the message.

Reading citations

Each citation links directly to the source page. Click any [n] reference or the URL in the sources list to open the page in a new tab. Always verify important facts from the original source — the AI may misquote or misattribute.

Limitations

• Paywalled content — Hubrix receives only the public snippet from each result, not the full article text.
• Very recent events — search results depend on Serper's index freshness, which may lag minutes to hours behind breaking news.
• Accuracy — the model synthesises multiple sources and can make mistakes. Cross-check critical information.
• Language — web search works best in English; results in other languages may be less consistent.

When to use web search