# What is Co-Browsing?

Also called collaborative browsing, cobrowse.

Co-browsing is a session in which a helper views, and sometimes interacts with, the same web page a visitor is on, scoped to that page rather than the visitor's whole device. Implementations either mirror the page structure to reconstruct it remotely or stream pixels of the page region. Field masking is a standard requirement so sensitive values never leave the visitor's browser.

The dominant approach is DOM-based. A script on the page serializes the document and streams subsequent mutations, so the helper's browser reconstructs an equivalent view. This is bandwidth-efficient, keeps text selectable, and allows selective redaction before transmission. Pixel-based approaches capture rendered frames instead, which is simpler for complex canvas or plugin content but heavier and less controllable.

Scope is the defining difference from screen sharing. A co-browse session shows only the page where the script runs, so other tabs, applications, notifications, and desktop content are never exposed. Sessions are started by explicit visitor consent, commonly by sharing a short code or accepting a prompt, and the visitor can end the session at any moment.

Masking rules are configured in advance rather than left to judgment during a session. Elements matching selected rules, typically payment fields, national identifiers, passwords, and free-text notes, are stripped or replaced before the stream leaves the browser, so the values are never present on the receiving side even in transit or in a recording.

Interaction levels vary from view-only, through annotation and pointer highlighting, to remote control where the helper can click and type. Each step increases both usefulness and risk, so many deployments allow annotation freely and require an additional per-session confirmation before control is granted, with all actions attributed in an audit log.

## Key points

- Scoped to one page, unlike full screen sharing
- DOM mirroring is lighter and allows selective redaction
- Masking rules strip sensitive fields before transmission
- Requires explicit consent and visitor-controlled termination
- Control levels range from view-only to remote interaction

## In practice

A visitor stuck on a multi-step application form accepts a co-browse invitation from a chat conversation. The helper sees the form as the visitor scrolls, with the identification number and payment fields rendered as blocked placeholders. The helper highlights the field causing the validation error rather than typing into it, the visitor corrects it, and the session ends automatically when the visitor navigates away from the form.

## Related terms

- [Chat Widget](/en/glossary/chat-widget)
- [Embedded Assistant](/en/glossary/embedded-assistant)
- [Deep Link](/en/glossary/deep-link)
- [Conversation Thread](/en/glossary/conversation-thread)

[Back to the AI Glossary](/en/glossary)