Intro
A citation map is one of the six baseline outputs of a Capston Core audit. It records, for each prompt and engine pair, the URL or URLs that the engine used as a source.
Read carelessly, it is a long list of links. Read carefully, it tells a premium brand which intermediaries currently own its narrative, which engines bypass the brand site entirely, and where a single source is leaking the wrong frame to every model.
This page explains what a citation map records, the five patterns that matter, and the order in which to act on them.
What a citation map records
A citation map is a structured table, not a screenshot.
For each row, the minimum columns are:
- Prompt — the exact query run, verbatim
- Engine — ChatGPT, Perplexity, Gemini, Google AI Overviews, others
- Model version — the build behind the answer on the capture date
- Capture date — the day and time the answer was recorded
- Citation URL — the source link the engine surfaced
- Citation domain class — own site, OTA, travel media, review site, destination board, partner page, outdated listing
- Citation position — order of appearance among cited sources
- Brand mentioned — yes or no, in the cited document
That structure is what makes a citation map auditable. Each row links back to the AI answer evidence layer, so a claim about source authority is traceable to an exact capture, not a recollection.
A citation map without domain class is a link dump. The class column is what turns a list into a diagnostic.
The five patterns to look for
Read the map for five things, in this order.
-
Own-domain citation rate. What share of prompts cite a URL on your own domain? A low rate means the brand site is not the source the model trusts on its own subject. This is the headline number — track it per engine, because engines differ widely.
-
OTA capture exposure. What share of prompts cite OTAs, aggregators, or marketplaces as the primary source? When an OTA is the cited source, the next click in the user journey routes commercial intent through that intermediary, not your direct channel.
-
No-citation prompts. Which prompts produced an answer with no citation at all? The model spoke without sourcing — pure parametric memory. These are the prompts where corrections, refreshes, and counter-evidence are hardest to inject, and where fact-accuracy risk is highest.
-
Repeat-domain framing. Which non-brand domains appear again and again across the prompt set? A travel magazine, a single review, a destination board page can become the de facto framing source for the brand across engines. If the framing is off, the leak is concentrated and fixable.
-
Position trends. When your own domain is cited, is it first, second, or buried? Position interacts with whether the engine quotes the source or merely lists it. First-position own-domain citations are the strongest signal that the model treats the brand site as authoritative on the prompt.
These five patterns map directly to the dimensions used in AI visibility scoring, so a citation map is not a parallel exercise — it is the evidence base for the score.
Domain class taxonomy
Every citation URL is tagged into one of seven classes. The taxonomy is deliberately narrow so the map stays comparable across audits.
- Own site — the brand’s primary domain and its language variants
- OTA — online travel agencies, aggregators, marketplaces selling the brand or its category
- Travel media — editorial publishers, magazines, guidebooks
- Review site — TripAdvisor, Google reviews surfaces, vertical review platforms
- Destination board — official tourism boards, regional or national
- Partner page — a page on a partner, distributor, or affiliate site about the brand
- Outdated listing — a directory or aggregator page with stale facts (closed venues, old rates, prior ownership)
The outdated-listing class is the one most brands underestimate. A single stale page can carry wrong opening hours or a discontinued offer into multiple engines for months. Citation maps surface these explicitly so they can be deindexed, corrected at source, or contested.
Reading order: from biggest risk to easiest win
Patterns matter only if they sequence action. The standard reading order is:
- Outdated-listing citations first. Wrong facts in a cited source contaminate every engine that reuses it. Fix at source.
- OTA capture on conversion-intent prompts. Where a booking-shaped query routes through an intermediary, the commercial leak is immediate.
- No-citation prompts on brand-defining topics. If the model speaks about the brand with no source, the brand has no leverage on the answer until evidence exists and is cited.
- Repeat-domain framing on comparison prompts. A single off-frame source carrying the brand across “best of” queries is concentrated risk and concentrated opportunity.
- Own-domain position improvement. Once the first four are stable, work the position of own-domain citations upward.
This order is not about effort — it is about which fix prevents the next one from being wasted. Refreshing brand content while an outdated OTA listing keeps getting cited is wasted refresh.
It is the same priority logic applied across the Capston Hospitality Scorecard: risk first, then leak, then gain.
How this fits into Capston Core
The citation map is one of six baseline outputs of the Capston Core methodology. It is the source layer underneath the score: citation share, source quality, and commercial risk all draw their evidence from the same captures the citation map records. Without the map, those scores are assertions. With it, they are traceable claims.
→ Back to Capston Core
FAQ
What is a citation map in AI search?
A structured record of which URLs each AI engine cited as a source, for each prompt in the audit set, with capture date and model version. It is the evidence base for source-authority analysis.
How is it different from a backlink report?
A backlink report counts links into the brand site. A citation map counts links out of AI engines — which sources the model chose to cite when answering a prompt about the brand. The two overlap only when the brand’s own pages are cited.
Can a brand influence which sources AI engines cite?
Indirectly. Fixing outdated listings, contesting wrong facts at source, publishing structured evidence on the brand domain, and earning citations from high-authority publishers all shift the set of sources the model trusts. The citation map tells the brand where to direct that effort.
How often should a citation map be refreshed?
Quarterly for most premium brands; monthly for high-stakes accounts or during a repositioning. Model updates and content changes both move the map.
Final CTA block
See which sources speak for your brand inside AI engines.
Get your citation map
Read the methodology