How to Read Your Sources Overview Stats in AmICited

Every answer ChatGPT, Perplexity, Gemini or Google AI Overviews gives is built on top of something: a set of web pages the model read, weighed and pulled facts from. The Sources overview stats are where you see that whole reading list summarized as six numbers.

At a Glance

• The Sources page stats bar shows six numbers: Citations, Domains, URLs cited, Your pages, External, New/wk.
• Compare Your pages against External to see how much of the AI answer is built from your own content.
• Citations, Domains, and URLs cited show volume, breadth, and concentration of the source landscape.
• New / wk is a freshness signal: a churning source pool means the window to earn a citation is still open.
• A high External-to-Your pages ratio usually points to a specific, fixable content gap.

What is an AI citation, and why do sources matter?

When an AI engine answers a prompt, it doesn’t invent facts from nothing. Modern answer engines retrieve a handful of web pages relevant to the query, read them, and generate a response grounded in that content, then, in many cases, attach a visible or inferable link back to the page they used. That link-back is an AI citation : a specific instance of a specific page being used as evidence for a specific answer. A brand can be mentioned in an AI answer without being cited, and it can be cited without being mentioned by name in the visible text, which is why citations and mentions are tracked as separate signals.

The pages that earn these citations are, collectively, “sources.” Some belong to the brand being discussed; most, in almost every category, belong to someone else: competitors, review sites, marketplaces, forums, news outlets, Wikipedia, documentation hubs. Which sources an engine picks, and how often it revisits them, is not random. Engines apply something close to AI source selection logic: they favor pages that are topically precise, structurally easy to parse, and judged trustworthy through a mix of domain reputation, citation history and independent corroboration, plus content freshness , since a page that was accurate a year ago but hasn’t been touched since is a weaker citation candidate than one updated last month.

This matters for a very practical reason: if you don’t know which domains and pages are shaping what AI says about your category, you’re optimizing blind. A brand that only watches its own website analytics has no visibility into the 80–90% of the answer that was built from other people’s pages. Understanding the source landscape (not just your own citation count, but the full set of domains AI engines trust for your topics) is a foundational part of generative engine optimization : the practice of making content structurally and topically easy for AI systems to find, trust and cite. The Sources overview stats are AmICited’s compressed view of that landscape.

Where the stats live

The Sources page shows every domain and page AI engines pulled from when answering your tracked prompts. The stats bar at the top summarizes that whole landscape in a handful of numbers: it’s the row directly under the Sources page title, which explains the page’s purpose: “Every domain and page AI engines pulled from when answering your tracked prompts, so you know who shapes what the models say about [your domain].”

What each stat means

• Citations: the total number of source citations across all your tracked prompts. This is a raw volume count: one prompt, answered by one engine, can generate several citations if the model draws on multiple pages to build its response. A prompt tracked across four engines multiplies that further. Citations is the broadest number in the bar and the one most sensitive to how many prompts and engines you’re monitoring.
• Domains: how many distinct domains have been cited. This collapses citation volume down to unique publishers. A domain with dozens of citations across many prompts still counts once here, which makes Domains a better read on the breadth of your competitive source landscape than Citations is.
• URLs cited: how many distinct pages, not domains, have been cited. A single domain can contribute many URLs, think of a competitor whose product page, comparison page and blog post are all cited separately. URLs cited sits between Citations and Domains: it tells you how concentrated or spread-out the citing is within each domain.
• Your pages: how many of those cited URLs are on your own domain. This is the number that answers “how much of the AI answer, for my topics, is actually coming from me?”
• External: how many cited URLs sit on other domains. Read together with Your pages, External shows you the split between owned and earned presence in AI answers, closer, conceptually, to earned media than to a traditional backlink count, because these are pages the AI model chose to read and use, not links other sites chose to point at you.
• New / wk: how many new sources appeared in the last week. This is a freshness signal for the source set itself: a topic where the citation pool keeps churning is one where AI engines are actively re-evaluating who to trust, which means the window to earn or lose a citation is still open.

How to use it

1. Gauge your footprint. Your pages ÷ URLs cited tells you what share of the cited web, for your topics, is actually yours. A low ratio is not automatically bad (some categories are inherently comparison-heavy, with review sites and marketplaces dominating regardless of how strong any one brand’s own content is), but it is a number worth tracking over time, and a shrinking ratio deserves attention.
2. Watch New / wk. A steady stream of new sources means the topic is active and the source set is worth revisiting regularly rather than treating it as settled. A flat New / wk over many weeks suggests the citation pool has stabilized around an established set of trusted domains, which makes it harder, but not impossible, to break in with a new page.
3. Set the scope. Like everything on the page, these numbers respect the top filters (AI model, country, date), so narrow them to focus on one engine or period. Citation behavior differs meaningfully between engines: the domains ChatGPT favors are not always the ones Perplexity or Google AI Overviews favor, so a single blended number can hide engine-specific patterns that matter for where you focus content or outreach efforts.
4. Cross-reference with rank. The stats bar tells you how much of the landscape exists; it doesn’t tell you where you rank within it. Pair what you see here with an AI rank tracker view of your own citation position to understand not just whether you’re present, but how prominently.

The stats are the summary; the charts and tables below them show which domains and pages make up the picture, letting you drill from the six headline numbers down to the individual URL that earned (or is denying you) a citation.

Turning the numbers into action

The overview stats are diagnostic, not decorative: each one points at a different lever. A high External number relative to Your pages usually means there’s a specific, findable gap: a comparison page, a “best of” list, or a forum thread that keeps outranking your own content for the same prompts. Once you’ve identified which external domains are doing the citing, the next step is usually content, not tracking: publishing pages that answer the same underlying questions more directly, with the kind of clear structure and factual density that make a page easy to extract from. If you’re not sure how to shape that content, a good next step is how to structure content so AI models can actually cite it , which walks through the formatting and framing choices that influence whether a page gets picked up at all.

A persistently low Domains count paired with a high Citations count can mean the opposite problem: a small handful of sources are doing all the work, which makes your AI visibility fragile: if one of those sources changes, de-indexes, or falls out of favor with a given model, your citation volume can drop sharply with it. Diversifying the source set you’re cited alongside, and increasing the number of domains that mention you favorably, is part of what separates a brand that owns how AI describes it from one that’s one algorithm update away from disappearing from AI answers. It’s also worth understanding how citation volume relates to competitive standing: two brands can have similar Citations counts and very different AI share of voice if one is consistently cited alongside strong competitors and the other isn’t cited in comparative contexts at all.

Finally, if the technical side of getting your own pages into the “Your pages” count is the bottleneck (not a content quality problem but a crawlability one), check whether your site is easy for AI crawlers to read in the first place; publishing an llms.txt file and confirming your key pages aren’t blocked is a low-effort way to remove a purely technical barrier to citation.

Where you go from the stats bar depends on what it’s telling you. If External dominates Your pages, the fix is usually a content and authority problem worth reading up on in depth: becoming the go-to source in your industry for AI covers the strategies for earning more of that citation share over time. If the split looks healthy but you want to know whether it’s improving or slipping relative to competitors, pair the Sources page with your rank and share-of-voice views on a recurring cadence, so a good week of New / wk turns into a lasting shift in who AI engines trust to answer your prompts, rather than a blip the stats bar happened to catch once.