browser agent
In plain wordsAn AI agent that uses a web browser on your behalf — searching, clicking links, filling in forms and reading pages — to complete tasks such as booking, shopping or gathering information.
An AI agentAI agent, n.: An AI system that does not just answer but acts: given a goal, it decides on steps, uses tools such as a browser or code editor, checks the results and keeps going until the job is done. that accomplishes tasks by operating a web browser: navigating to pages, reading content, clicking, typing into forms, scrolling and switching tabs. Browser agents observe pages through screenshots, the document object model (DOM), the accessibility tree or a mix, and act either through simulated input or through browser-automation interfaces such as the Chrome DevTools Protocol, Playwright or Selenium. They range from read-only research assistants to agents that log in, buy or submit forms for the user, and are increasingly built into browsers themselves. The open web makes them especially exposed to indirect prompt injectionindirect prompt injection, n.: A prompt-injection attack in which the trick instructions are hidden in content the AI reads on its own — a web page, an email, a file — rather than typed by the person using it. — hidden instructions on a page that hijack the agent — and to deceptive page designs, CAPTCHAs and irreversible transactions; mitigations include confirmation before sensitive actions, isolated or logged-out browser profiles, site allowlists and injection classifiers. Benchmarks include WebArenaWebArena, prop. n.: A test world of realistic, self-hosted websites — a code-hosting site, an online shop, a forum and more — in which AI agents must complete tasks such as finding orders or editing posts., Mind2Web and BrowseCompBrowseComp, prop. n.: A test of AI web-research agents: 1,266 puzzle-like questions whose answers take a long, persistent online hunt to uncover, though each answer is short and simple to confirm..
The browser agent got as far as the checkout page, then stopped to ask me to confirm the payment.
Origin
Research on web-navigating agents includes OpenAI's World of Bits platform and its MiniWoB tasks (2017) and WebGPTWebGPT, prop. n.: A 2021 OpenAI research system that answered questions by searching and browsing the web in a simple text browser, collecting sources and citing them — an early ancestor of today's research agents. (2021), which answered questions by browsing a text-only web. Consumer browser agents arrived in 2024–2025 with Google DeepMind's Project Mariner (December 2024), OpenAI's Operator (January 2025) and AI-first browsers that build an agent into the browser itself.
Usage note
Often used interchangeably with computer usecomputer use, n.: An AI's ability to work a computer through its normal screen and controls: it studies pictures of what is displayed and sends mouse and keyboard actions, so it can use ordinary apps that were built for people., but a browser agent is confined to the browser and frequently reads page structure directly instead of relying only on pixels.
Related
- computer usecomputer use, n.: An AI's ability to work a computer through its normal screen and controls: it studies pictures of what is displayed and sends mouse and keyboard actions, so it can use ordinary apps that were built for people.
- GUI agentGUI agent, n.: An AI agent that works through the buttons, menus and screens of ordinary software — on a desktop, phone or website — instead of through behind-the-scenes programming interfaces.
- WebArenaWebArena, prop. n.: A test world of realistic, self-hosted websites — a code-hosting site, an online shop, a forum and more — in which AI agents must complete tasks such as finding orders or editing posts.
- BrowseCompBrowseComp, prop. n.: A test of AI web-research agents: 1,266 puzzle-like questions whose answers take a long, persistent online hunt to uncover, though each answer is short and simple to confirm.
- indirect prompt injectionindirect prompt injection, n.: A prompt-injection attack in which the trick instructions are hidden in content the AI reads on its own — a web page, an email, a file — rather than typed by the person using it.
- WebGPTWebGPT, prop. n.: A 2021 OpenAI research system that answered questions by searching and browsing the web in a simple text browser, collecting sources and citing them — an early ancestor of today's research agents.
- deep researchdeep research, n.: An AI feature that spends minutes rather than seconds on a question — planning, running dozens of web searches, reading many sources — and then writes a long report with citations.
Sources
- Shi, T., Karpathy, A., Fan, L., Hernandez, J. & Liang, P. (2017). World of Bits: An Open-Domain Platform for Web-Based Agents. ICML 2017.
- Nakano, R., Hilton, J., Balaji, S. et al. (2021). WebGPT: Browser-assisted question-answering with human feedback. arXiv:2112.09332. arxiv.org/abs/2112.09332
Mentioned in
AI agentAI agent, n.: An AI system that does not just answer but acts: given a goal, it decides on steps, uses tools such as a browser or code editor, checks the results and keeps going until the job is done., computer usecomputer use, n.: An AI's ability to work a computer through its normal screen and controls: it studies pictures of what is displayed and sends mouse and keyboard actions, so it can use ordinary apps that were built for people., deep researchdeep research, n.: An AI feature that spends minutes rather than seconds on a question — planning, running dozens of web searches, reading many sources — and then writes a long report with citations., agentic browseragentic browser, n.: A web browser with an AI assistant built in that can do things for you online — reading pages, clicking, filling in forms, shopping or booking — instead of just showing you websites., agentic commerceagentic commerce, n.: Shopping and payments carried out by AI agents for people — finding products, comparing offers and completing purchases — within limits the person has approved., BrowseCompBrowseComp, prop. n.: A test of AI web-research agents: 1,266 puzzle-like questions whose answers take a long, persistent online hunt to uncover, though each answer is short and simple to confirm., CAPTCHACAPTCHA, n.: A small test on websites, like typing wavy letters or clicking every picture with a traffic light, meant to be easy for people but hard for computer programs, so as to keep automated bots out., GAIAGAIA, prop. n.: A test of AI assistants made of real-world questions that people can answer with a browser and some patience, but that require an AI to search, read files and combine facts., WebArenaWebArena, prop. n.: A test world of realistic, self-hosted websites — a code-hosting site, an online shop, a forum and more — in which AI agents must complete tasks such as finding orders or editing posts., GUI agentGUI agent, n.: An AI agent that works through the buttons, menus and screens of ordinary software — on a desktop, phone or website — instead of through behind-the-scenes programming interfaces., WebGPTWebGPT, prop. n.: A 2021 OpenAI research system that answered questions by searching and browsing the web in a simple text browser, collecting sources and citing them — an early ancestor of today's research agents.