How CMOs Should Set AI Search KPIs by Industry Instead of Using One Global Benchmark

Answer Capsule

CMOs should not turn one cross-industry AI citation average into a universal performance target. The July through September 2026 benchmark shows why.

Across 283 qualifying markets, the overall Top-100 source-portfolio quartiles were 57.48%, 63.93%, and 69.49% for persistence and 19.76%, 22.70%, and 25.79% for portability. Those numbers are useful as broad study context. They are not appropriate traffic-light thresholds for every industry.

When the 282 markets with named sector assignments were reclassified using their own detailed-sector quartiles instead of the global quartiles, 105 markets, 37.23%, changed persistence band; 132, 46.81%, changed portability band; and 108, 38.30%, changed gap band. Across the three measures, 211 of 282 markets, 74.82%, changed at least one four-band classification. 96 markets, 34.04%, changed at least two, and 38, 13.48%, changed all three.

The industry contrast can be extreme. In Legal Services, even the sector's 75th percentile is below the overall study's 25th percentile for persistence, portability, and the persistence-portability gap. In Property, Auto & Commercial Insurance, the sector's 25th percentile for portability is 26.40%, already above the overall study's 75th-percentile portability threshold of 25.79%.

That does not mean Legal should receive easier goals or Insurance should receive harder goals. It means these ecosystem metrics should be used as reference bands, not as universal brand-performance targets.

A defensible CMO scorecard should therefore separate two layers:

1. Company KPIs, such as recommendation coverage, cited-domain response presence, factual accuracy, platform coverage, and business outcomes, measured against the company's own declared baseline and priorities.
2. Market-context metrics, such as source persistence, portability, breadth, and depth, interpreted against the actual market first, then the detailed industry, then the macro sector, with the global benchmark as the broadest context.

This article develops the industry pillar, There Is No Universal GEO Strategy: How AI Citation Persistence and Portability Differ by Industry Across 283 Markets, and the benchmark-error analysis, Why Cross-Industry AI Citation Benchmarks Can Mislead CMOs.

The Overall Study Quartiles Are Context, Not Universal KPI Thresholds

Questions This Section Answers

1. Should a CMO use the 283-market persistence and portability quartiles as universal AI search targets?
2. What do the overall AI citation quartiles actually describe?
3. How should a CMO interpret a company's market when it sits above or below the global benchmark?

The benchmark measures source-portfolio structure at the market level. It does not measure one company's recommendation share, conversion rate, traffic, revenue, or marketing efficiency.

The overall 283-market reference points are:

A CMO can use those values to understand where a market sits in the full study. The mistake is turning them into a single corporate scorecard such as:

• below 57.48% persistence = bad,
• 57.48% to 63.93% = needs improvement,
• 63.93% to 69.49% = good,
• above 69.49% = excellent.

The data does not support those value judgments.

A market can fall below the global persistence quartile because its category has a more fragmented or changing source environment. A company can still be performing well inside that market. Conversely, a market can sit above the global portability quartile even though it is below the normal portability range for its own sector.

The benchmark should therefore answer, "What kind of source environment are we operating in?" It should not answer, "Is our company succeeding?"

Company success requires company-level measures.

Industry Quartiles Change the Context for Most Markets

Questions This Section Answers

1. How often does a market move to a different KPI reference band when industry quartiles replace global quartiles?
2. Which AI citation metric is most sensitive to using an industry-specific benchmark?
3. Is a macro-sector benchmark specific enough for executive reporting?

To test the KPI problem directly, each of the 282 named-sector markets was placed into one of four descriptive bands for persistence, portability, and the gap:

• below the first quartile,
• first quartile to median,
• median to third quartile,
• at or above the third quartile.

Those bands were calculated first using the overall 283-market distribution and then using the market's own detailed sector. All 19 preserved detailed sectors in this analysis contained at least 10 markets.

Portability is the most benchmark-sensitive of the three measures. Nearly half of the named-sector markets occupy a different four-band position when evaluated against their detailed sector instead of the global study.

The effect is even clearer across all three metrics simultaneously.

So 211 of 282 markets, 74.82%, change at least one band. More than one-third change at least two.

A macro sector improves the comparison, but it still does not eliminate the problem. Comparing macro-sector bands with detailed-sector bands, 150 of 282 markets, 53.19%, change at least one of the three classifications. 63 markets, 22.34%, change at least two, and 18, 6.38%, change all three.

That is the practical hierarchy for executive benchmarking:

actual market first, detailed industry second, macro sector third, global study last.

The broader levels remain useful for context. They simply should not replace the more relevant level when it exists.

Some Industries Sit Almost Entirely Outside the Global KPI Bands

Questions This Section Answers

1. Which industries are most distorted by a universal AI citation benchmark?
2. Can an entire industry's normal source environment sit outside the global interquartile range?
3. Why would the same portability percentage mean something different in Legal Services and commercial insurance?

The most striking evidence comes from comparing detailed-sector quartiles with the overall study quartiles.

For Legal Services, the sector reference bands are:

The Legal 75th percentile is below the overall 25th percentile on all three measures.

That means a market near the top of the Legal distribution can still look like a bottom-quartile market on a global dashboard.

The opposite pattern appears in Property, Auto & Commercial Insurance portability:

The insurance sector's 25th percentile is already higher than the overall study's 75th percentile.

A portability result of 25.79% therefore occupies opposite positions depending on the context. Globally it is at the upper-quartile boundary. In Property, Auto & Commercial Insurance it is below the sector's first quartile.

This is why industry benchmarking is not about granting different standards. It is about describing where a market sits inside the source environment that actually resembles it.

The Same Market Can Move From Bottom Quartile to Top Quartile Without Changing the Data

Questions This Section Answers

1. Can changing the comparison group move a market from the bottom to the top KPI band without changing the underlying result?
2. What are real examples of global AI citation bands reversing after industry normalization?

The underlying market value does not change when the comparison group changes. Only the interpretation changes.

Consider four examples.

Bankruptcy Lawyers is the clearest example. Its persistence of 52.67% is below the global first quartile of 57.48%, but it sits at the upper-quartile boundary for Legal Services. Its 35.02-point gap is also below the global first quartile but above the Legal Services third quartile of 34.09 points.

Boat Insurance reverses in the other direction. Its 25.79% portability reaches the overall study's third-quartile threshold, but it remains below the Property, Auto & Commercial Insurance first quartile of 26.40%.

Construction Management Software has 65.29% persistence. That looks above the overall median, but it is below the Business Software & Operations first quartile of 65.98%.

Online Therapy has 21.95% portability. Globally that sits between the first quartile and median. Within Behavioral Health, Treatment & Senior Care, it exceeds the sector third quartile of 21.77%.

None of these comparisons says the market is performing well or poorly. They demonstrate that relative position depends on the comparison population.

That matters when an executive dashboard uses colors, grades, or thresholds. The same number can change from red to green solely because the benchmark changed, even though nothing in the underlying market changed.

Industry-Specific Reference Bands for 19 Detailed Commercial Sectors

Questions This Section Answers

1. What persistence and portability ranges should CMOs use as industry context for different commercial sectors?
2. Which detailed sectors have materially higher or lower source-overlap reference bands?

The table below reports the observed middle 50% and median for all 19 preserved detailed sectors with at least 10 markets. These are reference bands, not recommended targets.

Several patterns are useful for KPI interpretation.

Insurance is not one uniform source environment. Property, Auto & Commercial Insurance has a portability median of 28.21%, while Life, Health & Personal Insurance is 23.84%. A single Finance benchmark compresses that difference.

Health is especially heterogeneous. Behavioral Health, Treatment & Senior Care has 50.38% median persistence, Medical, Dental & Clinical Services 57.48%, and Wellness, Beauty & Personal Health 66.67%.

Technology is also not one benchmark. Business Software & Operations has 69.49% median persistence, while Marketing, Sales, Communications & CX is 62.60% and Cybersecurity, IT & Web Infrastructure is 63.93%.

The detailed-sector bands should still be treated as context. A CMO operating in a specific category should continue down one more level to the actual market whenever sufficient data exists.

The KPI Scorecard Should Separate Company Performance From Market Structure

Questions This Section Answers

1. Which AI search metrics belong on a CMO scorecard, and which should be treated only as market context?
2. How should a CMO avoid confusing citation authority with recommendation performance?
3. What should be benchmarked against a company's own baseline versus an industry distribution?

A useful executive dashboard needs separate rows for company performance, source footprint, and market structure.

This distinction prevents a common category error.

If the source environment in a legal market has 50% persistence, that is not a target telling a law firm to reach 50% of recommendations. Persistence describes overlap between market source portfolios across observed months. Recommendation coverage describes how often the company is recommended in a declared prompt or response population.

The units are different. The denominators are different. The business meaning is different.

The same is true for portability. A 28% portability score in an insurance market says something about cross-platform resemblance of source portfolios. It does not mean the company has 28% visibility across platforms.

The upcoming Which AI Citation Metrics Belong on a CMO Dashboard? develops the cross-industry metric architecture. The industry-specific decision here is which context distribution should sit beside each metric.

Do Not Set the Sector Median as the Target

Questions This Section Answers

1. Should a CMO target the industry median for AI citation persistence or portability?
2. How should an AI visibility target be set when the sector benchmark is descriptive rather than prescriptive?

An industry median is not a performance objective merely because it is industry-specific.

There are at least four reasons.

First, persistence and portability describe the market source environment, not a controllable brand outcome. A company cannot reasonably be instructed to "increase market portability to 30%" as though portability were a brand-owned conversion rate.

Second, a higher value is not always a better commercial outcome. Higher persistence means more recurring source overlap across observed months. Higher portability means more source overlap across platforms. Neither proves higher recommendation share, better traffic, better leads, or stronger revenue.

Third, the study is observational. It does not establish what actions cause a market to move from one quartile to another.

Fourth, the company may have business priorities that make one platform, market, or buying intent materially more important than another. An unweighted industry median cannot choose those priorities.

A better target-setting sequence is:

1. Define the business outcome. Decide whether the program is trying to improve recommendation coverage, factual accuracy, source presence, publisher authority, or another declared outcome.
2. Lock the measurement unit. Specify markets, prompts or response sets, platforms, time window, geography, and denominators.
3. Measure the company baseline. This is the primary starting point for an improvement target.
4. Add the actual market context. Use source persistence, portability, depth, and platform-specific source structure to understand the environment.
5. Add the detailed-sector reference band. Use Q1, median, and Q3 to show where the market sits relative to commercially similar categories.
6. Show macro and global values only as broader context. They are valuable for orientation, not as the governing target.
7. Set the target from business priority plus baseline. Document the management choice instead of presenting it as a natural law of the benchmark.

This approach creates an industry-aware KPI system without pretending the research has established universal causal targets.

A Practical Executive Scorecard Can Use Two Columns of Truth

A CMO does not need a separate dashboard for every industry. The same dashboard architecture can work across markets if the benchmark context changes while the metric definitions remain fixed.

One practical format is to show two distinct columns beside every metric:

This keeps two truths visible at the same time:

• What happened to the company?
• What kind of source environment surrounds the company?

The market structure can help explain why a source strategy that works in one category does not transfer cleanly to another. It should not replace the company outcome.

That is the core reason industry-specific benchmarking belongs in the context layer of the KPI system rather than being treated as the KPI itself.

Methodology and Limits

The research uses the frozen July through September 2026 cohort of 283 qualifying tracked markets, 876,605 retained AI responses, and 8,718,629 citation occurrences resolved to registrable domains. The six surfaces are ChatGPT, Gemini, Perplexity, Microsoft Copilot, Google AI Overviews, and Google AI Mode. API copies and experimental surfaces are excluded.

Portfolio ranks count each domain once per retained response, sort by response presence, and break ties alphabetically. Eligible Top-100 comparison cells require at least 100 responses and 100 resolved domains. Thin or missing cells are excluded rather than assigned zero. Month portfolios pool platforms; platform portfolios pool completed months. Persistence and portability use Jaccard intersection divided by union, then the median eligible pair value within each market.

The overall full-study quartiles used in this article are calculated across the 283 qualifying markets. The detailed-sector band analysis uses the 282 markets with named sector assignments. The preserved taxonomy contains 19 detailed sectors with at least 10 markets, and all 282 named-sector markets fall into those groups. One unassigned market remains in the overall benchmark and is excluded from the industry-band analysis.

For each metric, a four-band context was created using Q1, median, and Q3. The same market was classified once against the full-study quartiles and once against its preserved detailed-sector quartiles. A band change means the market's relative position changed when the comparison population changed. It does not mean the underlying metric changed.

This analysis is descriptive. Sector samples are unequal and not randomized. Quartile bands are not validated performance grades, causal thresholds, or guaranteed target ranges. They describe the observed distribution of market-level source-portfolio metrics in this study.

The research is observational evidence from changing samples, not a certified identical-prompt panel. Of the retained responses, 104,071 used ingestion-date proxies. Only 246 markets had all three months and all six platforms eligible in the marginal portfolios. Shared source ecosystems, retrospective taxonomy, and candidate cross-job repetitions limit population inference. Domain presence does not verify ownership, article quality, recommendation influence, traffic, conversions, or revenue.

The proposed company-level KPI fields in this article, including recommendation coverage, factual inconsistency rate, and business outcomes, are measurement recommendations, not results reported by the frozen 283-market source-portfolio study.

Related Research

• There Is No Universal GEO Strategy: How AI Citation Persistence and Portability Differ by Industry Across 283 Markets
• Why Cross-Industry AI Citation Benchmarks Can Mislead CMOs
• How AI Citation Persistence and Portability Differ Across Finance, Software, Health, Legal, Home, and Consumer Markets
• How Publishers Can Choose Adjacent Verticals Based on Citation Stability and Platform Portability
• The Persistence-Portability Gap at Scale: 282 of 283 Markets Show Higher AI Citation Persistence Than Cross-Platform Portability Across 8.72 Million Citation Occurrences
• Can One GEO Strategy Work Across ChatGPT, Gemini, Perplexity, Copilot, Google AI Overviews, and Google AI Mode?
• Which AI Citation Metrics Belong on a CMO Dashboard?
• 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.>