Page Context Capture
Overview
This page explains how Ghast AI turns the page you are currently viewing into usable context, and how ordinary users should understand the boundary around that capability.
The main takeaway
For most users, three points matter most:
- Page context only works within the currently supported page scope.
- The capability can be enabled, reduced, or turned off through settings.
- Its purpose is to improve page-specific understanding, not to open unrestricted capture across the web.
What page context actually means
Page context simply means that Ghast AI does not look only at the question you type. On supported pages, it can also use what you are currently viewing to interpret your intent more accurately.
That means the goal is not to read as much as possible. The goal is to turn the current page into more usable context within a defined scope.
When it becomes active
This capability generally becomes active only when the following conditions are true:
- The current page is within the supported scope.
- The relevant page capability has not been turned off.
- Your current use actually depends on asking about the page.
If those conditions are not met, chat still works. It simply does not gain the same page-context enhancement.
How ordinary users should control it
This capability is best understood as a controllable page enhancement.
The more practical rule is simple:
- If you want Ghast AI to work around the current page, keep the relevant page capability enabled.
- If you want pure chat, or do not want the page to participate for the moment, turn that capability down or off.
What this boundary does not mean
- Do not treat page context as equally open on every website by default.
- Do not treat it as unconditional reading of everything on a page.
- Do not assume that weaker results on one page automatically mean something is broken; often the page is simply outside the supported range.
Ghast AI currently turns supported pages into usable context through its page-context capability. For most users, this should be understood as a bounded feature that improves page-specific understanding rather than a full-web capture model.