---
title: "Discovering images"
description: "How Blogged finds images on your website, the difference between scraping and capturing, and the limits on both."
updated: "2026-08-18"
verifiedAgainst:
  - packages/shared/src/assets.ts
---

Discovery reads your own site so your blog can be illustrated with your actual
product.

## Two mechanisms

<PropertyList>
  <Property name="Scraping">
    Reads the images referenced in your pages. Fast, and it finds what your site
    already publishes.
  </Property>
  <Property name="Capturing">
    Renders a page and captures it. This is what gets a usable picture of an
    interface that is drawn in the browser rather than served as a file.
  </Property>
</PropertyList>

## Limits

Discovery is bounded, and deliberately so. There are limits on how much is read
in one pass, and a cooldown between runs on the same site, so a repeated request
does not turn into pressure on your own servers.

Onboarding and manual runs use different budgets: the onboarding pass is the
broader of the two, because it is the one building the library from nothing.

## Duplicates

Images are deduplicated, so rediscovering the same picture on the next run does
not fill the review queue with copies of things you already decided about.

## What is not discovered

Anything behind a login, anything a crawler is asked not to read, and anything
that is not an image. If your best product visuals are inside the app, upload
them.