AI BEACON · CITATIONS
Every citation, traced back
to the page that earned it.
An answer that credits your domain tells you almost nothing. An answer that credits one guide, one thread or one review tells you what to write more of. AI Beacon reports the source at that level, and keeps a quoted source apart from one the model only read.
Credited by a model. Nobody knows which page.
Citation data that stops at the domain runs out of use in roughly three places.
They want to know which of their pages ChatGPT used. Naming the domain back to them is naming their own website, which they already knew.
You were credited in September and not in October. Without a record of the source there is nothing to compare, so the conversation turns into theories.
A three-year-old thread is a better source for your own product than your documentation is, and that only becomes actionable once you can see it.
Read the answer, then resolve the source.
A citation is only useful once it points at something you can open.
The prompt is answered
On each engine in turn, because the sources differ even when the answer looks similar.
Sources are taken from the answer
Everything the answer credits or draws on, in the order it gave them.
Each source becomes a page
A URL, a named Reddit thread or a YouTube video, rather than a domain.
Cited and consulted are separated
A source the answer quotes and a source it merely read are counted apart.
Four readings of the same citation.
Each one answers a question somebody is going to ask you.
The page, not the domain.
This is the detail that turns a citation into a decision. One page earned the place, and you can open it.
- A specific URLThe page the answer credited, as it credited it.
- Threads and videos tooA Reddit thread or a YouTube video is named at the same level of detail as a page.
- The domain alongsideAI Beacon shows the page as well as the domain, so both readings are there.
Quoted, or merely read.
Two very different things get called citations. Keeping them apart changes what you do next.
- CitedThe answer quotes or credits the source, which is what earns you the mention.
- ConsultedThe model read the source on the way to the answer without crediting it.
- Counted separatelyA page that is always consulted and never cited is a different problem from one that is never read at all.
Which of your URLs do the work.
Most sites have two or three pages earning nearly every citation, and a lot of pages earning none. Worth knowing which is which.
- Ranked by citationsYour own URLs ordered by how often answers credit them.
- Across prompts and enginesA page that works on one engine and not another is a useful signal.
- Pages never picked upThe ones that have never been read, which is where the technical checks come in.
The sources that are not yours.
Answers lean heavily on sources you do not own. Seeing which ones tells you where a mention is worth earning.
- Who gets creditedThe review sites, forums and channels your prompts keep drawing on.
- How oftenRepeat appearances rather than one-offs.
- What to do about itA source that keeps appearing is a link and a mention worth pursuing.
A domain is a clue. A page is an instruction.
The two readings lead to different work, which is why the level of detail matters more here than it looks like it should.
A typical stack
- –Citations recorded at domain level
- –Quoted and read sources counted as one thing
- –Screenshots as the record of a mention
- –No view of which of your own pages earn anything
With citation tracking
- ✓The page, thread or video named in full
- ✓Cited and consulted reported separately
- ✓A record per prompt and per engine, kept over time
- ✓Your own URLs ranked by the citations they earn
The rest of AI Beacon, and where this one sits.
Citations are the second-order signal. Presence in the answer comes first, and these pages cover both.
What is AI citation tracking?
AI citation tracking records which sources an AI answer credited or drew on, for each prompt you track. In AI Beacon every source is resolved to the page, Reddit thread or YouTube video rather than left at the domain, and a source the answer quotes is reported separately from one the model only read. Your own URLs are ranked by the citations they earn, across 7 AI engines plus Google AI Overview, AI Mode and Search.
Key facts
- A citation is reported as the specific page, thread or video, as well as the domain
- Cited sources and consulted sources are counted separately
- Your own pages are ranked by how often answers credit them
- Sources are recorded per prompt and per engine, and kept over time
- Brand presence inside the answer remains the primary signal; citations are second-order
- A citation is not a promise of a visit, and some AI traffic arrives with no referrer
Last reviewed 15 October 2026 by the SEORCE team.
What people ask about sources.
The page. A citation is reported as the specific URL, Reddit thread or YouTube video the answer used, with the domain shown alongside it.
A cited source is one the answer quotes or credits. A consulted source is one the model read without crediting it. They are counted separately, because a page that is read but never credited needs different work from one that is never read.
No. A citation puts your brand in front of the reader, and whether they click is another matter. Visits from AI tools are recorded in Web Analytics, and the attribution page reads them against your prompts, though some arrive with nothing attached to say where they came from.
Usually because the thread answers the question more directly, or because your page cannot be read properly. Open the prompt in prompt diagnosis to see which of your URLs were read, then check the page in Technical Suite if it was skipped.
Yes, on all seven plus Google AI Overview, AI Mode and Search. Each engine is read separately, because the same answer is often built from different sources on each one.
Yes. Build a dashboard and export it as Excel, PDF or Google Sheets, put it on a schedule, or, on Agency plans and above, read it through the SEORCE API.
See which page earned the citation.
Add a site, track a handful of prompts, and open the sources behind the answers you get.