Citation Intelligence

See the evidence behind the recommendations.

Citation Intelligence maps the pages and domains that appear around AI recommendations, then compares that evidence with your own coverage.

The workflow

Figure 1 — the Citation Intelligence loop
  1. 01Prompt
  2. 02AI answer
  3. 03Citation
  4. 04Source classification
  5. 05Competitor comparison
  6. 06Evidence gap
  7. 07Action
  8. 08Remeasurement

The loop closes: the last step is the first question set, rerun.

  1. 01

    Prompt

    A fixed buyer or seller question from the locked market set.

  2. 02

    AI answer

    The response captured verbatim, with provider and timestamp.

  3. 03

    Citation

    Every cited URL and domain recorded against that answer.

  4. 04

    Source classification

    Each source assigned to a class: portal, brokerage, publication, review platform, forum, social, own site, institutional.

  5. 05

    Competitor comparison

    Which sources repeatedly appear around competitors, and whether you appear on them.

  6. 06

    Evidence gap

    Where comparable coverage is absent, inaccurate or fragmented.

  7. 07

    Action

    A legitimate, documented correction or addition to the evidence environment.

  8. 08

    Remeasurement

    The same question set rerun, with the change log alongside the result.

We do not reverse-engineer a private algorithm. We reverse-engineer the observable evidence environment around AI recommendations.

What an engagement delivers

Seven artifacts, each traceable to preserved answers you can read.

Citation inventory
Every cited URL behind the captured answers, with its class, date and accuracy check.
Recurring source analysis
The classes and domains that appear repeatedly around recommendations in your market.
Competitor source map
Where competitors are represented and where you are absent on the same sources.
Entity clarity report
Name variants, duplicate records and conflicting facts found across public sources.
Citation-gap matrix
Each gap scored on importance, frequency, confidence, difficulty and relevance.
Source-priority plan
A tiered, documented sequence of legitimate corrections and additions.
Remeasurement report
The same question set rerun after the change window, with the change log beside the result.

Citation inventory

For every captured answer, we record the following fields. The inventory is the raw material for everything else on this page.

  • Cited URL
  • Cited domain
  • Citation order within the answer
  • Source title
  • Publication date
  • Source type
  • Recommended entity in that answer
  • Whether the citation supports the recommendation
  • Whether the cited information is accurate

Recurring source analysis

Each cited source is assigned to a class so patterns can be compared across markets rather than domain by domain. The classes that recur are where evidence work pays.

  • Listing portals
  • Brokerage websites
  • Local publications
  • Industry publications
  • Review platforms
  • Community forums
  • Social and video platforms
  • Team websites
  • Government or institutional sources
Anonymized measured example · Jersey City, NJ · indexed frequency, most-cited class = 100
Source classIndexed frequency
Listing portals100
Brokerage sites44
Local publications22
Community forums12
Video7
The team's own site0

Observed association, not causal proof. Indexed frequency expresses how often each class appeared around recommendations relative to the most-cited class in the same captured set.

Measured fact

Competitor source mapping

The source map shows which sources repeatedly appear behind competitors’ recommendations and whether the client has comparable coverage on those same sources. It is an observational comparison of evidence footprints — not a statement about why any particular answer was produced.

Measured factInterpretationHypothesisRecommendation

Every line in a delivered report carries one of these labels, so a count is never confused with an explanation.

Example structure — values are populated from the client's own measured data
Evidence patternRecommended teamOverlooked team
Local publication mentionsMeasured valueMeasured value
Brokerage profile citationsMeasured valueMeasured value
Listing portal coverageMeasured valueMeasured value
Own website citedMeasured valueMeasured value
Entity naming consistencyObservedObserved

Placeholder structure. No values are shown until they come from an approved measurement.

Entity clarity analysis

Name variants
Every form of the team name found across public sources.
Duplicate profiles
Multiple competing records for the same team or agent.
Stale brokerage affiliations
Profiles still listing a former brokerage.
Conflicting production claims
Different volume or transaction figures across sources.
Incorrect locations
Service areas or offices stated wrongly.
Fragmented team-member information
Members attributed inconsistently or not at all.

Citation-gap matrix

Importance
How central the gap is to the questions the client cares about.
Frequency
How often the pattern appears across captured answers.
Confidence
How strongly the observation is supported by the preserved evidence.
Difficulty
Effort, cost, dependencies and time to close the gap.
Relevance
Fit with the client's target market and target question set.

What this cannot prove

  • A source appearing around a recommendation does not prove it caused the recommendation.
  • Citation sets vary between runs; the pattern across answers is the finding, not any single citation.
  • Closing a gap does not guarantee a citation, a recommendation or a position.

Source-priority recommendations

Gaps are sequenced by what can be corrected honestly and quickly, not by what sounds impressive.

Tier 1 — correct what is wrong
Inaccurate or conflicting facts on sources already cited in your market. Lowest effort, highest confidence, no new coverage required.
Tier 2 — complete what is thin
Sources cited around competitors where you have a record but it is incomplete, outdated or fragmented.
Tier 3 — earn what is missing
Classes of source that recur around recommendations in your market and carry no record of your team at all. Slowest, and dependent on third parties.
Never
Hidden text, prompt injection, fabricated reviews, manufactured forum accounts, mass-generated pages, or any claim your production cannot substantiate.
Recommendation

Before-and-after remeasurement

After the change window, the same question set is rerun under the same conditions and reported beside the change log.

Remeasurement report structure — populated only from the client's own measured runs
MeasureBaseline runRemeasured runChange
Recommendation shareMeasured valueMeasured valueReported
Citation shareMeasured valueMeasured valueReported
Own site citedMeasured valueMeasured valueReported
Entity inconsistencies openMeasured valueMeasured valueReported
Documented actions in windowChange logListed

Placeholder structure. No before-and-after figures are published here; a remeasurement is reported to the client from their own runs, and published only with written permission.

What this cannot prove

  • A change between runs is not proof that an action caused it; answer sets vary between runs on their own.
  • The change log records what was done, not what the model did with it.
  • No remeasurement outcome is promised, guaranteed or sold in advance.

How to read these findings

We do not claim that a citation proves causation. We identify recurring evidence patterns worth investigating and test whether legitimate improvements change the measured result.

Request a citation analysis for your market.

One client per market. Start by telling me where you work.