Which AI Citation Metrics Belong on a CMO Dashboard?

Answer Capsule

A CMO dashboard should not contain every AI visibility metric that can be calculated. It should contain the smallest set of measures that answer different management questions without hiding denominators or collapsing source authority, recommendation authority, factual accuracy, and business outcomes into one score.

The July through September 2026 LLM Authority Index evidence shows that some citation metrics are meaningfully distinct while others are largely diagnostic restatements of the same underlying source structure.

Across 283 qualifying markets:

• persistence and portability had a Spearman correlation of only 0.352, so monthly source recurrence and cross-platform source resemblance should be shown separately;
• the persistence-portability gap correlated 0.859 with persistence but only -0.111 with portability, which means the gap is useful as a summary but can obscure which axis is actually changing;
• in the 260 complete six-platform markets, portability correlated 0.871 with the share of Top-100 sources appearing on at least four platforms, 0.868 with the all-six-platform source share, and -0.870 with the size of the six-platform Top-100 union;
• therefore, portability belongs on the executive dashboard, while 4+ platform share, all-six share, union size, and single-platform share belong mainly in drill-downs that explain why portability is high or low.

The same non-redundancy appears at the domain level. Among 2,743 domains that reached a Top 10 at least once:

• Top-100 market breadth and Top-10 market breadth correlated 0.608, strong enough to be related but far from interchangeable;
• Top-10 market breadth and same-market three-platform Top-10 repetition correlated 0.430;
• Top-10 market breadth and same-market all-six repetition correlated only 0.229;
• Top-10 market breadth and source-specific monthly recurrence correlated -0.090, effectively showing that broad depth and temporal recurrence answer different questions.

The counting unit matters too. The frozen benchmark contains 8,718,629 resolved citation occurrences but 6,825,821 response-domain appearances. Citation occurrences are 27.73% higher because the same domain can be cited multiple times within a response. Occurrences describe citation activity. Response-domain appearances are closer to source reach. They should not be used interchangeably.

A defensible executive dashboard should therefore separate four layers:

1. Company outcome: recommendation coverage, factual consistency, and attributed business outcomes.
2. Company source footprint: response presence, source depth, market breadth, same-market platform coverage, and recurrence for the company-owned domain or relevant third-party sources.
3. Market context: persistence and portability, interpreted against the actual market and relevant industry.
4. Diagnostics: citation occurrences, platform-specific source maps, single-platform share, all-six share, category concentration, URL-level evidence, and other explanatory fields.

The frozen source study directly supports the source and market-structure layers. It does not directly measure company recommendation coverage, sentiment, conversions, or revenue. Those require separate datasets.

This article develops the metric framework from What Makes a Domain an AI Citation Authority?, the industry KPI guidance in How CMOs Should Set AI Search KPIs by Industry Instead of Using One Global Benchmark, and the publisher proof standard in How Publishers Can Prove AI Citation Authority to Advertisers Without Promising Rankings.

Key Findings

A CMO Dashboard Needs Different Metrics for Different Questions

Questions This Section Answers

1. Which AI visibility questions should an executive dashboard answer separately?
2. Why should recommendation, citation, accuracy, and business metrics not be blended into one score?
3. Which measurements come from source data and which require separate company-level data?

A useful dashboard starts with questions rather than available fields.

A CMO usually needs to know at least four things:

Are buyers seeing the company?

That is a recommendation or mention question. It requires a declared prompt or response population and a company-level coding protocol.

What sources are AI systems using around the company and category?

That is a citation-source question. The frozen 283-market research directly measures this layer at the registrable-domain portfolio level.

Are the answers factually accurate and commercially safe?

That is an answer-quality and inconsistency question. It requires reviewed answer content, not just citation portfolios.

Is any of this producing business value?

That requires traffic, referral, lead, conversion, pipeline, or revenue attribution outside the citation study.

A single "AI visibility score" can conceal movement in opposite directions.

A company's recommendation coverage can rise while its owned domain disappears as a cited source. A publisher's source authority can increase while the associated brand is never recommended. A company can be recommended more often while factual inconsistency worsens. Citation activity can increase while qualified traffic stays flat.

Those are not contradictions. They are different measurement layers.

The dashboard should make the layers visible instead of forcing them into one number.

Citation Occurrences and Response Presence Are Not the Same KPI

Questions This Section Answers

1. Should a CMO use raw citation counts as the primary AI source metric?
2. What is the difference between citation occurrences and response-domain appearances?
3. Why can citation activity increase without an equivalent increase in source reach?

The benchmark makes this distinction measurable.

Across the frozen cohort there were:

• 8,718,629 citation occurrences resolved to registrable domains;
• 6,825,821 response-domain appearances.

The occurrence count is 1.277 times the response-domain appearance count, or 27.73% higher.

That difference exists because a domain can be cited more than once inside the same response.

If one answer cites the same publisher three times, the occurrence count can rise by three while response presence for that publisher in the answer rises by one.

Neither metric is wrong.

They answer different questions.

For executive reporting, response presence is usually the cleaner reach metric because the denominator can be stated directly.

Occurrences remain useful in the diagnostic layer. They can show whether a source is being cited repeatedly inside answers, but they should not be labeled unique reach.

A dashboard that reports only occurrences risks rewarding repetition within a smaller response set as though it were broader distribution.

Persistence and Portability Both Belong Because They Are Not Redundant

Questions This Section Answers

1. Are monthly AI citation persistence and cross-platform portability measuring the same source behavior?
2. Should a CMO report both persistence and portability on the same dashboard?
3. Why can a market be stable over time while remaining fragmented across AI platforms?

Across all 283 qualifying markets, the Spearman correlation between persistence and portability was only 0.352.

That is enough to show some relationship, but not enough to treat one metric as a substitute for the other.

Persistence asks:

How much do the observed source portfolios recur across completed months?

Portability asks:

How similar are the observed source portfolios across AI platforms?

A market can be stable month to month but still have very different source sets on ChatGPT, Gemini, Perplexity, Copilot, Google AI Overviews, and Google AI Mode.

That is the central pattern in the corpus. Median persistence across the 283 markets was 63.93%, while median portability was only 22.70%.

The two metrics therefore deserve separate executive rows.

The dashboard should also preserve their units and comparison populations. Persistence is a median of eligible monthly portfolio-pair overlaps. Portability is a median of eligible platform-pair overlaps. A value of 60% on one axis and 25% on the other does not represent the same type of denominator.

The Persistence-Portability Gap Is a Summary, Not a Third Independent KPI

Questions This Section Answers

1. Should the persistence-portability gap be treated as a standalone executive KPI?
2. What information can the gap hide if persistence and portability are not shown beside it?

The gap is useful because it compresses a common pattern into one number.

But it is mathematically derived:

gap = persistence - portability

That makes it different from an independent measurement.

Across 283 markets, the gap had a Spearman correlation of 0.859 with persistence and only -0.111 with portability.

This does not make the gap invalid. It means the observed variation in the gap is much more closely aligned with variation in persistence than with portability in this cohort.

A dashboard that shows only the gap can therefore hide materially different market structures.

For example:

• 65% persistence and 25% portability produces a 40-point gap;
• 50% persistence and 10% portability also produces a 40-point gap.

The strategic environments are not the same.

The executive design should therefore show:

Persistence | Portability | Gap

with the gap treated as a convenience field or alert, not as a replacement for the two underlying metrics.

This follows the same principle used in the industry KPI research: a derived benchmark should not erase its components.

Portability Does Not Need Four More Executive Metrics Beside It

Questions This Section Answers

1. Which cross-platform citation metrics are useful diagnostics but redundant on the executive dashboard?
2. How closely do single-platform share, all-six share, and source-union size track portability?
3. What should move from the CMO scorecard into a platform drill-down?

The 260 complete six-platform markets allow a direct redundancy test.

For each market, the source analysis measured:

• total six-platform Top-100 union size;
• share of sources appearing on only one platform;
• share appearing on four or more platforms;
• share appearing on all six;
• and the market's published portability value.

The correlations are strong:

These fields are useful because they make portability concrete.

A market with a large union and high single-platform share has a fragmented source environment. A market with a larger 4+ platform core and all-six core has more reusable sources across platforms.

But putting portability, union size, 4+ share, all-six share, and single-platform share as five equal executive KPIs creates a dashboard that repeatedly reports the same underlying structure.

A cleaner architecture is:

Executive row: portability

Drill-down:

• six-platform source-union size;
• single-platform share;
• 4+ platform share;
• all-six source count/share;
• pairwise platform overlap matrix;
• platform-specific unique-source counts.

That preserves diagnostic detail without treating every explanatory field as a separate corporate KPI.

Breadth, Depth, Platform Repetition, and Recurrence Are Different Enough to Keep

Questions This Section Answers

1. Which publisher or source-authority metrics add independent information to a CMO dashboard?
2. Can market breadth substitute for Top-10 depth, cross-platform repetition, or recurrence?
3. Why should broad authority and durable authority be reported separately?

The domain-level evidence provides the opposite result from the portability drill-down.

Among the 2,743 domains that reached a Top 10 at least once, the authority dimensions are related but not interchangeable.

The implications are practical.

Market breadth and depth both matter

A domain can appear in many Top-100 markets while rarely reaching Top 10. The commercial authority work found 116 domains with Top-100 breadth of at least 50 markets, including 46 that converted fewer than 10% of those markets into Top-10 depth.

A CMO assessing source authority should therefore keep breadth and depth separate.

Platform breadth and same-market repetition both matter

A domain can reach Top 10 on multiple AI systems somewhere in the study without those platforms overlapping in the same market.

The dashboard should prioritize same-market platform coverage for the advertiser's actual category rather than global platform breadth alone.

Recurrence deserves its own row

The near-zero Spearman relationship between Top-10 breadth and recurrence, -0.090, is especially important.

A source can be broad but unstable. Another can be narrow but recur reliably in the markets where it appears.

For publisher and source reporting, breadth and recurrence answer different management questions:

• breadth: where does the authority exist?
• recurrence: how consistently does the source remain visible across observed months?

Both deserve separate fields when the CMO is evaluating a publisher, owned domain, or third-party source strategy.

Category Concentration Is Context, Not a Universal KPI

Among the 2,743 Top-10 domains, Top-10 market breadth and concentration in the domain's largest macro sector had a Spearman correlation of -0.555.

Broad domains tend to be less concentrated, while many specialist authorities earn their depth inside a narrower category family.

That does not make category concentration a metric that every CEO needs on the front page.

It is more useful when the business question is:

• Is this publisher relevant to our category?
• Is this source authority generalist or specialist?
• Does a prospective distribution partner's global footprint actually come from our sector?
• Is the company expanding into adjacent markets where its authority already transfers?

For those decisions, category concentration belongs in the source or publisher drill-down.

For general executive reporting, the actual market result should remain primary.

Recommended Executive Dashboard Architecture

A CMO dashboard can remain compact if it separates executive KPIs, market context, and diagnostics.

Layer 1: Company outcome KPIs

These require company-level data outside the frozen source study.

These are not citation-source metrics and should not be inferred from them.

Layer 2: Company and publisher source-footprint KPIs

These metrics are distinct enough to retain separately.

Layer 3: Market-context metrics

The industry benchmark should follow the hierarchy established in the P5 research: actual market first, detailed sector second, macro sector third, global study last.

Layer 4: Diagnostic drill-downs

These explain the executive metrics rather than competing with them:

• citation occurrences;
• platform-pair overlap matrix;
• six-platform source-union size;
• single-platform source share;
• 4+ platform source share;
• all-six source count/share;
• platform-specific unique sources;
• cited URL inventory;
• category concentration;
• source type;
• month-by-month and platform-by-platform rank tables;
• collection changes and data-quality flags.

This structure keeps the front page small while preserving enough evidence for analysts to explain changes.

What Should Not Be a Standalone CMO KPI

Several numbers are useful but become misleading when elevated without context.

Raw citation occurrences

Useful for activity. Poor as a stand-alone reach metric because repeated citations inside one response can inflate the count relative to response presence.

One global authority rank

Useful for discovery. Weak for decision-making because category concentration, source depth, platform repetition, and recurrence can differ substantially among similarly broad domains.

Persistence-portability gap alone

Useful summary. Insufficient without the two underlying axes.

Six-platform breadth somewhere in the corpus

Useful descriptive fact. Does not prove same-market six-platform authority.

One composite AI visibility score

Potentially convenient, but the frozen research does not establish a validated weighting scheme that combines recommendation coverage, citation source depth, recurrence, factual accuracy, and business outcomes.

Any such score is a management model. It should be labeled as one and expose its component metrics.

Citation authority as a proxy for recommendation authority

Not supported. The final P6 article addresses this distinction directly: AI Citation Authority vs AI Recommendation Authority: Why Being a Source Is Not the Same as Being Recommended.

A Useful Dashboard Preserves the Denominator and the Nulls

The dashboard should make every metric auditable.

For each result, store:

• market;
• platform;
• period;
• response count;
• eligible or ineligible status;
• source-depth cutoff;
• domain or URL identity;
• numerator;
• denominator;
• collection or methodology changes;
• and periods where the desired result did not occur.

Thin or missing cells should remain missing or ineligible. They should never be silently converted to zero.

This matters especially in trend reporting.

If a platform had no eligible Top-100 portfolio in one period, the dashboard should not show the source as having "zero authority" for that period. If the collection scope changes, the chart should annotate the change rather than presenting the movement as pure performance.

A good executive dashboard is compact because it chooses a small number of non-redundant metrics, not because it discards the audit trail.

Methodology and Limits

This analysis uses the frozen July through September 2026 LLM Authority Index cohort of 283 qualifying tracked markets, 876,605 retained AI responses, 8,718,629 citation occurrences resolved to registrable domains, and 6,825,821 response-domain appearances across ChatGPT, Gemini, Perplexity, Microsoft Copilot, Google AI Overviews, and Google AI Mode.

Market-level persistence and portability are the published median eligible pairwise Jaccard values for month portfolios and platform portfolios respectively. The persistence-portability gap is persistence minus portability.

The 260-market cross-platform diagnostic analysis uses markets with all six eligible Top-100 platform portfolios. For each market, the six-platform source union and source platform-breadth counts were calculated from those strict portfolios. Single-platform share, four-plus-platform share, and all-six share are percentages of the market's six-platform Top-100 union.

The domain-level analysis uses the same 1,655 eligible market-platform Top-100 portfolios. Top-100 and Top-10 market breadth count a market once when a domain reaches the relevant cutoff on at least one eligible platform. Same-market platform repetition counts markets where the domain reaches Top-10 depth on at least three or all six surfaces inside the same market.

Source-specific recurrence uses the 246 markets with all three months and all six platforms eligible. For each domain, the denominator is the balanced markets where the domain entered at least one pooled monthly Top-100 portfolio. The numerator is the subset where the domain appeared in the pooled Top 100 in all three months.

Correlation results in this article use Spearman rank correlation unless otherwise stated. Correlation is descriptive and does not establish causation. High correlation does not prove two metrics are identical, and low correlation does not prove they are statistically independent under every population.

The source study does not directly measure company recommendation coverage, sentiment, factual inconsistency rates, referral traffic, leads, conversions, pipeline, or revenue. Those fields in the proposed dashboard architecture are measurement recommendations that require separate data.

The research is observational evidence from changing samples, not a certified identical-prompt panel. Of the retained responses, 104,071 used ingestion-date proxies. Shared source ecosystems, retrospective taxonomy, and candidate cross-job repetitions limit population inference.

Citation is not recommendation. Domain presence does not verify publisher ownership, editorial independence, page quality, sponsorship status, recommendation influence, traffic, conversions, or revenue.

No metric in this article should be interpreted as a guaranteed future ranking or a causal estimate of marketing performance.

Related Research

• The 2026 AI Commercial Citation Authority Index
• What Makes a Domain an AI Citation Authority? A Practical Measurement Framework
• How CMOs Should Set AI Search KPIs by Industry Instead of Using One Global Benchmark
• How Publishers Can Prove AI Citation Authority to Advertisers Without Promising Rankings
• Citation Breadth vs Citation Depth: Which AI Authority Metric Should CMOs Care About?
• Can One GEO Strategy Work Across ChatGPT, Gemini, Perplexity, Copilot, Google AI Overviews, and Google AI Mode?
• AI Citation Authority vs AI Recommendation Authority: Why Being a Source Is Not the Same as Being Recommended

The earlier Persistence-Portability framework provides conceptual context. The 2026 AI Citation Authority Study covers an earlier high-stakes sample and is a separate population.