Web search

AI Web Search With Sources You Can Open

Training data has a cutoff. The web does not.

Every model has a knowledge cutoff, and the failure is not that it says "I do not know" — it is that it answers confidently from memory that is eighteen months stale. Clade's web search tool, backed by Perplexity Sonar, pulls live results into the conversation with linked citations, so the answer is current and checkable.

When to turn search on

Anything with a date attached — news, releases, prices, policy changes.
Library and framework documentation, which moves faster than any training run.
Company, product and competitor facts that you will repeat in public.
Statistics and figures you intend to cite.
Anything where being wrong would be embarrassing rather than merely inconvenient.

The advantage of search as a step, not a destination

Dedicated AI search engines answer the question and stop. That is fine until you need to do something with the answer, at which point you are copying findings into another tool that knows nothing about where they came from.

On Clade, search returns into a thread that also contains 50+ reasoning models, file upload, code execution and image generation. The model that found the sources drafts the memo from them, and can go back for more mid-draft.

Getting better results out of it

Ask specific questions. "Latest AI news" returns noise; "what changed in the EU AI Act in the last six months" returns sources.
Include a timeframe when recency matters.
Ask for multiple viewpoints on contested topics rather than the consensus.
Open the citations. Presence of a link is not proof it supports the claim.
Follow with a reasoning model to turn results into a brief or a decision.

What it will not fix

Bad sources. Search finds what is published, including things that are wrong.
Paywalled and gated content, which stays out of reach.
Very recent events in the first hours, before anything credible is indexed.
Your own judgement about which source deserves weight.

Models you can use for this

MiMo-V2-Flash256K

DeepSeek-V3.2 (Reasoning)128K

DeepSeek-V3.2128K

Qwen3-Max256K

Qwen3.5-Plus1M

Command A Reasoning256K

Command A256K

MiniMax M2197K

See all 8+ models and specs

Frequently asked questions

Does Clade search the live web?

Yes. Web search is a built-in tool that pulls live results into the conversation with linked citations, powered by Perplexity Sonar models.

Are the citations real?

They are live links you can open, which is a meaningful step above a model reciting a reference from memory. Open them anyway — a real link attached to a claim it does not support is still possible.

Is this the same as Perplexity?

It uses the same Sonar models for search, but as a tool inside a broader workspace rather than as the whole product. You can search, then reason with GPT or Claude, then run code, in one thread.

Do I need to turn search on manually?

You can enable it for a conversation when a question depends on current information. Leaving it off is cheaper in credits for questions that do not.

Does search cost extra?

It draws from the same monthly credit allowance as everything else — no separate subscription or per-search fee.

Keep reading

Current answers, checkable sources

Live search plus 50+ reasoning models in one thread. Start on the free tier.