Web Search on Magicdoor: Fact-Checking Guide

Fact-check with current web search

Current Magicdoor access: Magicdoor uses an Exa-backed web search tool on supported chat models, not a Perplexity search integration. The older Perplexity Reasoning and Deep Research (pplx) picker models are retired. Those older choices remain a historical part of this guide, not a current invitation to select them. Perplexity's own separate product lineup is unaffected; check Magicdoor's current models.

For a question that depends on fresh facts, choose a search-enabled model and enable web search. Ask for dated original URLs, open the sources, and check what they actually say. A citation is a lead, not a guarantee: compare primary documents and confirm publication dates before relying on a claim. When one answer is insufficient, ask a narrower follow-up or switch to a supported model for an independent synthesis. Web search has a separate provider-reported usage cost in addition to the chat model's usage; compare current costs.

For practical workflows, see how to use web search for research and how to use multiple AI models.

Historical context

Older versions of this page recommended Perplexity Reasoning for focused web-backed answers and Deep Research (pplx) for longer multi-source investigations. They were separate Magicdoor chat picker entries, not currently selectable. Today's search tool is not a promise of identical output or research depth. Preserve old conversations, but select a current supported model for new research.

FAQs

How can I use web search on magicdoor.ai?

Choose a current search-enabled chat model and use web search when the task needs fresh sources. Verify original URLs; search is a tool, not a separate chat model.

Which Perplexity chat models are currently selectable?

None. Perplexity Reasoning and Deep Research were older Magicdoor picker entries and are now retired. Web search remains available with supported current chat models.

Can I use search results with other models?

Yes. Verify the source URLs, then switch to another supported model in the same conversation for synthesis. Model changes may affect tool and attachment support.