# Sources

When an AI engine answers a buyer's question, it leans on real web pages. The **Sources** page shows you which ones — the domains and URLs the engines pulled in while talking about your category. It's how you turn "AI keeps recommending a competitor" into "because it keeps citing these three pages", and from there into a content plan. For the precise definition of each metric, see the [Metrics reference](/docs/metrics).

## Retrieved vs cited

Two things can happen to a page inside a single answer, and Voxoria tracks them separately:

- **Retrieved** — the engine _fetched_ the page while working out its answer. It shaped the response even if the reader never sees it. This is the engine's research surface.
- **Cited** — the page appears as a _visible source or footnote_ in the answer the reader sees. This is visible attribution.

The two are **independent**. A page can be retrieved without being cited (it informed the answer but wasn't surfaced), and some engines cite a page without exposing that they fetched it. Keeping them apart preserves a real distinction: _what the engine read_ versus _what the reader is shown as the source_. Retrieval metrics measure influence; citation metrics measure credit.

## Domains and URLs

The page opens on two tabs:

- **Domains** — sources rolled up by site (e.g. `g2.com`, your own domain, a competitor's).
- **URLs** — the individual pages.

A date-range filter applies to both. On Domains you can also filter by **domain type** (news, blog, review, official, and so on); on URLs you can filter to a single domain to see just its pages.

### The Domains tab

Two charts sit above the table: **Source retrieval by domain** plots how often the top domains were pulled in over your window, and **Sources type** breaks retrievals down by the kind of site. The table below lists each domain with:

- **Domain type** — the category of site.
- **Retrieved** — the share of answers in which any page from the domain appeared as a source.
- **Retrieval rate** — the average number of pages from that domain pulled into a single answer. Above 1.0 means engines typically pull several pages from the site at once.
- **Citation rate** — how often the domain is cited per answer in which it's used.

Click a domain to jump to the URLs tab filtered to its pages.

### The URLs tab

The same chart pairing — **Source retrieval by URLs** and a **Sources type** breakdown by page type (homepage, article, listicle, comparison, and so on) — sits above a table of individual pages:

- **Type** — the structural classification of the page.
- **Mentions** — which of your tracked brands are named on the page.
- **Retrievals** — how many times the page appeared as a source in your window.
- **Citation rate** — how often it's cited per answer that uses it.
- **Updated** — when Voxoria last read the page's content.

## The URL detail page

Click any URL to open its detail page — the full story of how one page shows up across your answers. It carries:

- A **KPI strip** — retrievals, citation rate, brands mentioned on the page, and when it was last read.
- **Retrievals over time** — the page's pull-in count across your window.
- **Retrievals by channel** — which engines lean on the page, and how often.
- **Mentions** — the tracked brands named on the page itself.
- **Prompts** — which of your tracked questions surface this page.
- **AI answers** — the individual captured answers that retrieved or cited it.

This is where you decide what to do about a page: a competitor's comparison article that gets retrieved across half your prompts is a content gap worth closing; a review site cited everywhere is a placement worth pursuing. To turn that read into a prioritised plan, run a [GEO audit](/docs/audits).