Browser agent
Definition
A browser agent is an AI agent that uses a web browser on your behalf: it reads pages, clicks, fills in forms and moves between tabs to complete a task. It can take over web chores you would otherwise click through by hand.
By Ronan Liedmeier, fractional head of growth and course author · Published 17 October 2026 · Last updated 17 October 2026
Also called: AI browser, browser-use agent
Comparing prices across five sites or copying details from a dozen pages becomes one request. The cost is a new kind of risk, because every page the agent reads was written by someone else.
How it works
The agent looks at the page, as text, a screenshot or both, decides on an action, takes it and looks at the result. It can come as an AI-native browser, a browser extension, or an agent that controls a browser remotely.
Because the agent acts on what it reads, text on a page can carry instructions aimed at it. Anthropic's research on browser use gives the example of an email with hidden white text that the user never sees but the agent processes. This attack is called prompt injection, and the same article states that no browser agent is immune to it.
Example
An ecommerce manager asks a browser agent to check the listed price and stock status of her top ten products on four competitor sites and put them in a table. Her instructions say read only, no logins, no forms. Before taking the table into her pricing meeting, she checks two rows against the live pages.
Common mistake
The risky setup is a browser agent that holds saved logins or payment details and submits forms without approval. Keep passwords and purchases with a person, and require an approval step before any action that submits or changes something on a site.
Learn it in AI Academy
- AI in the BrowserAdvanced · 14 min
Related terms
- AI agent
- An AI agent is an AI system that works toward a goal by choosing its own next steps: it picks a tool, uses it, looks at the result and repeats until the task is done. You set the goal and the limits, and the agent works out the path.
- Prompt injection
- Prompt injection is an attack where text an AI reads, such as an email, web page or file, contains instructions that make the model do something its user did not intend. OWASP ranks it first in its Top 10 risks for LLM applications.
- Human in the loop
- Human in the loop is a setup where an AI system pauses at set points so a person can approve or correct its work before it takes effect. It is used for actions that are hard to undo, such as sending emails to customers or moving money.
- AI guardrails
- AI guardrails are checks and limits placed around an AI system that block or flag unsafe or wrong inputs and outputs before they cause harm. The strongest ones run as code outside the model, so a cleverly worded request can't talk its way past them.