Web Search
Pencarian Web
A tool that lets the model look things up instead of guessing.
Web search as a tool gives a model the ability to look up facts that are not in its training data, including anything that happened after its knowledge cutoff. It is the cheapest way to close the gap between what the model knows and what is true today.
It also introduces new problems. The retrieved page is untrusted content that arrives in the same context as the instructions, which is a prompt injection surface. Search results are commercial, ordered, and often SEO-manipulated, so a confident answer can be built on a page that exists to sell something.
The good pattern: search, then read the source, then cite it. The bad pattern: search, then answer from the snippet. Snippets are written to be clicked, not to be accurate.
Related terms
RAG
Fetch the relevant documents first, then let the model answer using them.
Grounding
Tying the answer to something checkable, so it can be verified instead of trusted.
Knowledge Cutoff
The date after which the model has no trained knowledge at all.
Tool
A function the model is allowed to call, like reading a file or running a command.
Prompt Injection
Text that tries to make the model ignore its instructions and follow the attacker's instead.
Want this working inside your team?
It starts with one conversation about your workflow.