result.design_system.brand_assets to give your agents visual references from a company’s campaigns and social channels. Each image has a hosted URL, a source, a category, and a short description of its visible content and design.
Full extractions include this section by default. You can also extract it on its own. It returns still images, including video thumbnails. The separate assets section contains logos and media extracted from the website itself.
Extract brand assets
Submit a fresh extraction with onlybrand_assets selected:
submission_id until status is completed or failed, following the submission lifecycle. On completion, read result.design_system.brand_assets.assets.
Omit sections to include brand assets alongside colors, typography, the brand profile, and the rest of the design system. You do not need to provide social handles or credentials for the source platforms.
Brand assets are available from fresh extractions. Existing results are not updated automatically. Set
force: true to bypass an older cached result. A full extraction that already contains brand assets can satisfy a later selective request. See Caching and freshness.Read the result
The following example shows thebrand_assets section. The image URL is illustrative.
url points to the stored image, not the original post or ad. The response does not include source-page links, judge scores, or rejected images.
Categories
Candidates are reviewed for brand ownership and useful visual design before they are stored. Reposts, screenshots of customer posts, unrelated companies’ advertising, and incidental photos are excluded. The verified X account’s avatar and header are kept when available and downloadable, without passing through visual judging.
Sources
The extractor reads social links and metadata across the submitted page to identify company accounts. X, LinkedIn, and Instagram Highlights require a matching account on that page. Meta ads require a Facebook page link. If an account is missing, that source is skipped; Google Ads, email campaigns, and billboards can still produce results using the company’s name and domain.
An extraction returns up to 24 images, with at most six per source. The number varies with account discovery, available source content, and visual review. A result does not need to contain images from every source.

