Set up Prism curation enrichment
Prism is StackUp's curation enrichment engine. It reads and understands your content, then uses that understanding to enrich curated demand with real contextual signals. For Prism to work, it needs a way to discover your articles. This guide covers the two paths: automatic discovery through the AI Ad Ops SDK, or connecting a content source such as your sitemap.
What Prism does
Prism scans your pages and builds a content understanding layer on top of them: what each article is about, how pages relate to each other, and which contextual attributes describe your inventory. That layer is what powers curation enrichment, so demand can be matched to your content on substance rather than on URL patterns or coarse site-level categories.
The quality of enrichment depends directly on coverage: the more of your content Prism can discover and scan, the better it represents your inventory.
Already running the AI Ad Ops SDK? You can skip setup
SDK-based discovery has a useful property: it naturally prioritizes the content that actually receives traffic, so Prism's understanding is always weighted toward the inventory you are actively monetizing. If you have not installed the SDK yet, start with Set up the AI Ad Ops SDK, and Prism will follow along on its own.
The rest of this guide is for publishers who want to run Prism without the SDK, or who want to guarantee coverage of content that receives little traffic.
Connect a content source
Without the SDK, Prism needs an explicit source that tells it which URLs to read. In the console, go to Curation enrichment › Content sources and add one of the following:
| Source | How it works | Best for |
|---|---|---|
| Sitemap | Point Prism at your XML sitemap (or sitemap index). Prism fetches it on a schedule and scans new and updated URLs. | Most publishers. The recommended default, since sitemaps already cover your full article inventory. |
| RSS / Atom feed | Prism polls your feed and scans new items as they publish. | News and high-frequency publishing, where fresh articles should be enriched quickly. |
| URL list | Upload or paste a list of URLs for Prism to scan. | Fixed sections, evergreen libraries, or a controlled test on a subset of pages. |
Sources are composable. A common pattern is a sitemap for full coverage plus an RSS feed so breaking content is enriched within minutes rather than on the next sitemap pass.
Setup steps
Verify your domain. Prism only scans verified domains. If you completed domain verification during SDK setup, the same verification applies here.
Add the content source. Enter your sitemap URL (for example https://example.com/sitemap.xml), feed URL, or URL list. Prism validates the source immediately and reports how many URLs it found.
Start the initial scan. Prism performs a first pass over the discovered URLs to build its baseline understanding of your content. Large sites are scanned progressively, so enrichment becomes available as coverage grows rather than waiting for a full crawl.
Let it run. After the initial scan, Prism re-checks your sources on a schedule, picking up new articles and re-scanning updated ones. No maintenance is required unless your sitemap or feed URL changes.
Verify enrichment is working
Once scanning has started, the Curation enrichment area of the console shows discovery and scan progress for each source: URLs found, pages scanned, and when the source was last checked. You will know Prism is healthy when:
- Your content source shows a recent successful fetch.
- Scanned page counts are growing toward the number of discovered URLs.
- New articles appear as scanned shortly after they publish.
Troubleshooting
- Source validation fails: confirm the URL is publicly reachable and returns valid XML (for sitemaps and feeds). Sources behind logins or IP allowlists cannot be scanned.
- Coverage stops growing: check whether your sitemap actually lists the missing URLs. Section-only sitemaps are the most common cause of partial coverage.
- Running both the SDK and a sitemap: that is fine. Prism merges discovery from all sources and scans each page once, so there is no duplication penalty. The sitemap simply guarantees coverage of low-traffic pages the SDK may not see.