# What gets captured

> Automatic text extraction, and drawing the area yourself when a layout defeats it.

Source: https://cognipage.io/docs/capture-areas

Once a source is approved, CogniPage has to decide which part of the window is the thing you are reading. Almost always it works this out by itself. This page is about the setting that governs that, and what to do on the pages where automatic extraction cannot win.

**Where** Settings → Tracking → What gets captured

## Find the text automatically

The default. CogniPage picks out the article and skips menus, sidebars, navigation, cookie banners and comments, then keeps the paragraphs you dwell on rather than everything on the page. This is right for nearly every web page, document and reader.

## Let me draw the area

Some layouts defeat any automatic reading: multi-column academic PDFs, dense course pages, readers that draw text to a canvas. For those, you can point at the text instead.

1. **Switch the setting on**
   In *Settings → Tracking*, choose **Let me draw the area (guided)**.
2. **Draw a box the first time each source is tracked**
   The screen dims and you drag a rectangle around the text you are reading. Press `Esc` to cancel.
3. **Read normally**
   Only what is inside the box is captured. The box is remembered for that page or document, so you draw it once, not every session.

## Redrawing an area

Layouts move. Press `⌘` `⇧` `T` (macOS) or `Ctrl` `⇧` `T` (Windows) at any moment to draw a new area for the page in front of you — you do not have to be in guided mode to use it, and it takes effect immediately.

Every area you have drawn is listed under **Authored** in *Settings → Tracking*, where you can clear any of them and go back to automatic text extraction for that source.

## The outline

**Outline the area while capturing** keeps a thin border around the region being read, so you can always see what CogniPage is looking at. It never blocks clicks or typing. It is on by default; turn it off for a completely clear screen.

> **Which should you use?**
>
> Leave it on automatic. Reach for the guided mode only when a specific document is producing junk — headers, page numbers, the wrong column — and use `⌘` `⇧` `T` as a one-off fix rather than switching modes globally.

## Windows that publish no text at all

Scanned PDFs, image-only documents and some browsers expose nothing readable to the accessibility layer. Where [Screen Recording](https://cognipage.io/docs/permissions) is granted, CogniPage recognises the text from the pixels instead, on your device. The image is read and discarded in memory — no screenshot is written to disk or transmitted. Without that permission, those windows simply produce nothing.