What is the difference between sources and citations in AI search?
source domains, retrieved sources, used vs cited
Sources are every page an AI engine retrieves while building an answer. Citations are the subset of those pages it shows to the reader as links. A page can influence an answer without being cited, so tracking only visible citations underestimates which domains shape what AI says about your brand.
Sources are all pages an engine retrieved; citations are the ones it displays.
Uncited sources can still shape the wording and recommendations in an answer.
Tracking both gives a fuller map of which domains influence AI answers.
How sources and citations work in AI answers
Sources and citations come from the same retrieval step but end up in different places. When an engine answers a question, it often splits the query into several searches, a process called query fan-out. It then retrieves pages for each search and uses them to ground the answer.
All retrieved pages are sources. The engine then chooses some of them to display as links. Those displayed links are citations. The rest stay hidden, even though the engine may have used their content.
For example, an engine might retrieve 20 pages for a product comparison question and show only 5 as citations. The other 15 still informed the answer.
Why the difference between sources and citations matters
The difference between sources and citations matters because brands that only track visible citations miss a large share of the pages that shape AI answers. A review site that is retrieved often but rarely cited can still decide which brands get recommended.
Looking at sources as well as citations helps you:
Find influential domains that never appear as visible links
Understand why an answer favors a competitor when the citations look neutral
Prioritize outreach by real influence, not only by visible credit
See whether your own pages are being read even when they are not shown
Both views support better decisions about AI citations and content.
How to track sources and citations
You track sources and citations by recording both the full set of pages an engine retrieved and the links it displayed, for each tracked prompt and run. Where an engine exposes its retrieved pages, those can be logged directly. Displayed citations are captured from the answer itself.
Run the same prompts across engines on a schedule.
Log retrieved source URLs and cited URLs separately.
Group URLs by domain and by source type, such as editorial or UGC.
Compare how often each domain is used versus cited.
A domain with a high used-to-cited ratio is a quiet influencer worth pursuing. This view builds on retrieval-augmented generation.
How KIME tracks sources and citations
KIME's citation tracking separates the source domains engines rely on from the links they show, for every tracked prompt. Each domain is typed, for example Editorial, UGC or Influencer, so you can see which kinds of sites shape answers and which ones get visible credit. In Track & Analyze you can compare your source footprint with competitors and spot influential domains to target with content or outreach.
Frequently asked questions
Can a page influence an AI answer without being linked?
Yes. AI engines often retrieve more pages than they display. A page that was read but not shown as a link can still shape the facts, wording and brand recommendations in the answer. That is why tracking only visible citations gives an incomplete picture of which sites influence AI answers in your category.
Why do AI engines cite fewer pages than they read?
AI engines cite fewer pages than they read to keep answers readable and to credit the most relevant sources. The exact selection logic is not public and differs by engine. In practice, some retrieved pages provide background or confirm facts, while only a few are shown as links to the reader.
What is a used-to-cited ratio?
A used-to-cited ratio compares how often a domain appears among the pages an engine retrieved with how often it appears as a visible citation. For example, a domain used 40 times and cited 8 times has a ratio of 5 to 1. A high ratio suggests strong hidden influence on AI answers.