Skip to main content

​
What are sources and citations?

Sources are the links an AI search engine attaches to its answer to back up what it says. When a model like ChatGPT or Perplexity searches the web, it cites the pages it used: as numbered inline references like [1] and [2] in the answer text, or as source cards showing the URL and page title. Promptwatch records each of those cited URLs as a citation, tied to the response, the prompt, and the AI model that produced it. Citations only exist where the model actually searched the web. A response with zero citations is normal for models answering purely from training data; Citations vs. mentions covers which platforms return sources and why a citation is not the same thing as your brand being discussed.

​
The three kinds of source in a response

Open a response and its sources are grouped by how the model used them (responses collected without this classification show one flat source list instead):
  • Inline Citations: sources the AI cited directly inside the response text. These carry the most weight, because they back specific claims.
  • Footnotes: sources listed in the sources footnote at the end of the response. Cited, but not linked to a specific sentence.
  • Search Results: pages the AI browsed while searching but did not cite in the answer. Useful to see what influenced the response without making it into the sources.
The distinction matters when you read a single answer: a page in the inline tier shaped what the model said, a page in the search-results tier merely got looked at.

​
What Promptwatch records per citation

Every citation carries the data the reports are built from:
  • The URL and its domain: so citations aggregate both per page and per site.
  • Position in the source list: counted from 1, because the sources a model lists first are the ones it leaned on hardest. This feeds citation rank.
  • Source type: the kind of site the page lives on, from news outlets to Reddit threads. The full list is in Source types.
  • Content type: what the page is, such as a listicle, review, or comparison.
  • Domain rank (DR): how authoritative the root domain is; a higher number means a harder domain to compete with.

​
Reading the Citations page

Open Citations in the sidebar. The filters at the top (models, topics, tags, prompt types, date range, and Hide ignored brands, on by default) apply across the tabs; the competitors filter applies on All Citations and URL Positions, and Citation Trends always analyzes its trailing 90-day window regardless of the start date. Each monitor also has its own citations view scoped to that monitor’s prompts.
  • All Citations: the Top Cited Domains chart shows the most-cited domains over time, with a line and bar toggle, next to a Citation share list with each domain’s share of all citations. Your own domain is always in the list, even at zero citations; its share is your self-citation rate. The table below lists every cited URL with its Source, Type, Models, Citations, Avg. position, DR (domain rank), and Best and Worst position.
  • URL Positions: charts where the most-cited URLs land in source lists over time, with a toggle between Position (where the URL ranks) and Rate (what share of responses cite it). This covers all domains, so you can watch which competitor pages hold the top slots. The metric itself is explained in Citation rank.
  • Citation Trends: the lifecycle view. It analyzes the trailing 90 days and shows, per page, when it was first cited, a 90-day sparkline (Citation rate), how many days it took to peak (Ramp up), its highest weekly citation share (Peak Citation), and how many days after the peak its weekly rate halved (Decay rate, empty while it hasn’t). Charts above the table break citation rate down by type and citation share by domain rank. This is where you see that citations are earned and lost over time, not held forever.
Within each tab you can narrow the table further by URL search, source type, content type, and domains (include or exclude). All Citations and URL Positions also export the filtered result as CSV or Excel.

​
Your pages vs. everyone else’s

The same citation data answers two different questions, depending on whose domain you look at:
  • Direct citations, links to your own website: are models building answers from your content? That share is the self-citation rate, and where those citations land is your citation rank.
  • Indirect citations, external sites that mention your brand: which pages you don’t own feed the answers in your category? Promptwatch scans those cited pages for brand appearances, which becomes Offsite Mentions, and two platform sources get their own pages: Reddit and YouTube.

​
Acting on what you see

This page is the reading guide; the plays live in the playbooks. If third-party platforms dominate your citations, start with the Reddit and YouTube strategy. If your own pages should be cited but aren’t, work through Why isn’t my page cited? and Optimize content. If models can’t fetch your pages at all, crawler logs show you where discovery breaks before citations ever start.