A concrete example
I turn on web_search in OpenAI's API. Now the model can go look things up live on the internet for its answers.
Why it matters
Web Search breaks through the model's knowledge cutoff. You get real-time info without coding your own scraper.
You run into it in OpenAI, Gemini with Google Search, ChatGPT Search, Claude Search, and news-watching agents.
Don't mix it up with
Tool use / Function calling: Tool use lets a model call an external tool instead of only replying with text.
Deep Research: Deep Research runs a long search, cites its sources, and hands back a structured report.
Common mistakes
- Thinking Web Search removes hallucinations (sometimes it adds them).
- Forgetting that searches cost extra on top of tokens.
- Confusing OpenAI's Web Search with Perplexity's Sonar (they're different).
Quick checklist
- First I check whether the word names a concept, a tool, a risk, or a metric.
- I tie it to a concrete case: I turn on web_search in OpenAI's API. Now the model can go look things up live on the internet for its answers.
- I keep the main trap in mind: Thinking Web Search removes hallucinations (sometimes it adds them).
Quick questions
What is Web Search tool in AI?
The Web Search tool lets the model go check the web before it answers.
Where will I run into Web Search tool?
You run into it in OpenAI, Gemini with Google Search, ChatGPT Search, Claude Search, and news-watching agents.
Which word should I read next?
Start with Tool use / Function calling, Deep Research, Responses API.