Browsing mode
Browsing mode is the LLM-product capability of fetching live web pages during a conversation — distinct from a single search tool call, browsing involves navigating multiple pages, clicking through, and extracting content as the model decides.
Browsing mode sits between a basic search-tool call and full [[deep-research-mode]]: the model navigates web pages live during the turn, fetching, scrolling, following links, extracting content as needed. ChatGPT's browsing mode, Claude's browser tool, Gemini's browsing, Perplexity's Pro Search all implement variations. Engineering challenges: latency (page fetches add seconds), bot-blocking (Cloudflare / hCaptcha block headless fetchers), content extraction (JS-rendered pages, paywalls), prompt injection (malicious page text instructs the model). Browsing mode is the right primitive for sub-minute queries needing 1-5 live page reads; deep research is the right primitive for multi-source longer-form synthesis.
When to use browsing mode
- Queries needing live (post-cutoff) content.
- Specific URL lookups (read this page).
Common mistakes
- Browsing without prompt-injection defense — malicious pages can hijack the agent.
- Browsing when a search tool call would suffice — slower + more expensive.
FAQ
What is browsing mode?
Browsing mode is the LLM-product capability of fetching live web pages during a conversation — distinct from a single search tool call, browsing involves navigating multiple pages, clicking through, and extracting content as the model decides.
When should I use browsing mode?
Queries needing live (post-cutoff) content. Specific URL lookups (read this page).
What are the most common mistakes with browsing mode?
Browsing without prompt-injection defense — malicious pages can hijack the agent. Browsing when a search tool call would suffice — slower + more expensive.
Related terms
- Browser agent — A browser agent is an LLM-driven system that controls a real or headless web browser to navigate sites, fill forms, click, and extract data — automating tasks that require interacting with web UIs.
- Web retrieval tool — A web retrieval tool is the agent-callable API that fetches live web content for LLMs — Tavily, Exa, Serper, Brave Search API, Perplexity API are 2026 leaders, used as the 'web search' function in tool-use loops.
- Computer use — Computer use is the agent capability where an LLM controls a real desktop or browser via screenshots + mouse/keyboard primitives — Anthropic introduced it in 2024 and it's mainstream across Claude, GPT, Gemini in 2026.
Last updated: 2026-06-01. Raw markdown: https://promtable.com/glossary/browsing-mode.md.