Answer Capsule

Across 283 qualifying tracked commercial markets, AI citation-source portfolios were dramatically more stable from month to month than they were consistent from one AI platform to another. The July through September 2026 benchmark includes 876,605 retained AI responses and 8,718,629 citation occurrences resolved to registrable domains across ChatGPT, Gemini, Perplexity, Microsoft Copilot, Google AI Overviews, and Google AI Mode.

The median market had:

• 63.93% citation persistence across eligible month pairs
• 22.70% citation portability across eligible platform pairs
• a 39.68 percentage-point Persistence-Portability Gap

Persistence exceeded portability in 282 of 283 markets, or 99.65%.

The pattern was not driven by a small number of extreme categories. The middle 50% of markets had gaps between 35.22 and 45.99 percentage points. 245 markets, 86.57%, had gaps larger than 30 points, and 136 markets, 48.06%, had gaps larger than 40 points.

The result also survived major descriptive sensitivity checks. At the Top 25 source level, the median gap widened to 46.99 points. At Top 50 it was 43.44 points. Excluding taxonomy concerns produced a 39.88-point median gap. Every one of the 183 markets with no ingestion-date proxy responses still had a positive gap, with a median of 40.29 points.

The practical implication is not that AI citation portfolios are static. They are not. It is that source authority appears substantially more durable across time than portable across AI systems. A publisher, source, or domain that repeatedly matters in one market may still fail to transfer across ChatGPT, Gemini, Perplexity, Copilot, Google AI Overviews, and Google AI Mode.

This is an observational comparison of changing stored samples. It does not establish identical prompts, identical model versions, a fixed retrieval panel, causation, future ranking performance, or recommendation influence.

Key Findings

Citation Persistence and Citation Portability Measure Different Things

Questions This Section Answers

1. What is the difference between AI citation persistence and cross-platform citation portability?
2. Can an AI source remain important over time without appearing consistently across ChatGPT, Gemini, Perplexity, Copilot, and Google AI?
3. Which citation metrics should a CMO track separately when measuring durable AI authority?

AI visibility reports often compress several forms of authority into one score. That can hide an important distinction.

A source can be:

• cited frequently,
• cited across many prompts,
• cited across many markets,
• cited by many AI platforms,
• stable over time,
• prominent within a market,
• or associated with company recommendations.

Those characteristics are related, but they are not interchangeable.

Citation persistence measures how much a market's important cited-domain portfolio recurs across time. In this study, each eligible month receives a ranked Top-100 domain portfolio. The month portfolios are compared using Jaccard overlap, and the market's persistence score is the median of its eligible month-pair comparisons.

Citation portability measures how much a market's important source portfolio transfers across AI platforms. Each eligible platform receives a Top-100 domain portfolio pooled across completed months. Those platform portfolios are compared using the same Jaccard method, and the market's portability score is the median of its eligible platform-pair comparisons.

The Persistence-Portability Gap is calculated within each market:

Persistence minus portability

A positive gap means the market's source portfolio was more similar from month to month than it was from platform to platform.

That distinction matters commercially. A domain can remain highly relevant to one market across multiple months while having limited coverage on another AI platform. In other words, durability does not imply universality.

The earlier Persistence-Portability framework introduced the conceptual distinction. This expanded benchmark tests it across a much larger set of commercial markets.

The Main Result Is Remarkably Broad: 282 of 283 Markets Had a Positive Gap

Questions This Section Answers

1. How consistently did AI citation persistence exceed cross-platform portability across commercial markets?
2. Was the Persistence-Portability Gap driven by a few extreme industries or visible across most of the 283-market benchmark?
3. How large was the typical difference between month-to-month citation stability and cross-platform source overlap?

The headline result is unusually one-sided.

Of the 283 qualifying markets, 282 had persistence greater than portability.

That is 99.65% of the benchmark.

Only one market had a negative value, and no market landed exactly at zero.

The median market recorded:

• 63.93% persistence
• 22.70% portability
• 39.68 points of difference

The mean gap was almost identical at 39.38 points, which is useful because it suggests the median was not hiding a radically different average market experience.

The distribution is even more informative than the headline median.

Only 12 markets, 4.24%, had gaps of 20 points or less. Only 38 markets, 13.43%, had gaps of 30 points or less.

The middle half of markets fell between 35.22 and 45.99 points. The descriptive 95% bootstrap interval around the median gap was 38.49 to 41.07 points.

This is not the profile of a result created by three or four extreme categories. The center of the distribution itself is large.

Persistence was also high in absolute terms. 251 of 283 markets, 88.69%, had persistence of at least 50%. Portability showed the opposite pattern: 274 of 283 markets, 96.82%, had portability of 30% or less.

That is the central empirical shape of the benchmark: most markets had a recurring temporal source core, while cross-platform source agreement remained much lower.

The Markets With the Largest Gaps Show How Extreme the Pattern Can Become

Questions This Section Answers

1. Which commercial markets had the largest gaps between AI citation persistence and cross-platform portability?
2. Can AI citation sources remain highly stable over time while being poorly portable across platforms?
3. Which markets provide the clearest examples of durable but platform-specific citation authority?

The largest gaps were not confined to one vertical family. They appeared across consumer products, insurance, health, beauty, pet care, mobility products, and business services.

These markets are useful because they demonstrate what the gap means in practice.

Long-Term Care Insurance had the highest persistence in the table at 85.19%, yet its median platform portability was only 23.46%. Outdoor Apparel and Technical Outfits showed an even larger absolute gap because its portability was below 20% despite persistence above 80%.

Life Insurance Companies presents another strong example. Its persistence was 72.41%, while portability was just 14.94%. A source strategy built from one platform in that market would therefore risk missing much of the citation environment observed elsewhere.

Stairlifts is commercially interesting because it combined a relatively large sample, 4,600 retained responses and 44,638 resolved citation occurrences, with a 55.04-point gap. This is not merely a tiny niche behaving erratically because of a handful of answers.

The same is true for Veterinary Pet Food and Prescription Pet Nutrition, which included 5,800 retained responses and 63,736 resolved citation occurrences and still produced a 54.96-point gap.

These examples reinforce the broader result: large persistence-portability gaps can exist in well-populated markets with substantial observed citation activity.

The Smallest Gaps Are Just as Important as the Largest Ones

Questions This Section Answers

1. Which markets had the smallest differences between citation persistence and portability?
2. Are there commercial markets where cross-platform AI citation overlap comes close to month-to-month stability?
3. Which markets should CMOs avoid benchmarking against the overall 39.68-point median?

A useful benchmark should preserve markets that do not look like the median.

The bottom of the distribution contains several important exceptions and near-exceptions.

These markets reveal two different ways to get a small gap.

One is low persistence and moderate portability, as in Cybersecurity Services and Financial Technology and Banking Software. Another is relatively high portability, as in Immigration Law Firms, where portability reached 39.86%, the maximum observed market-level portability in the 283-market cohort.

A CMO in Legal Services, banking software, or cybersecurity should therefore not assume that the 39.68-point overall median describes the local market. The correct benchmark is market-specific.

This is one reason the broader corpus includes There Is No Universal GEO Strategy: How AI Citation Persistence and Portability Differ by Industry Across 283 Markets.

Time Was Not Uniform: August to September Was More Stable Than July to September

Questions This Section Answers

1. How much did AI citation-source overlap change between July, August, and September 2026?
2. Was month-to-month citation persistence equally strong across every pair of months in the benchmark?
3. Which month pair showed the highest median AI citation-source persistence?

The overall persistence statistic summarizes eligible month-pair comparisons within each market, but the three calendar pairs were not identical.

The August-September comparison was materially more similar than July-September in the observed sample.

That is descriptive, not causal. The study cannot establish whether the difference came from market behavior, source changes, prompt mix, collection configuration, model changes, retrieval changes, or some combination of factors.

But the result is strategically important for measurement programs. A single month-to-month comparison can create a misleading impression of “normal” citation stability. July to September had a median overlap of 58.73%, while August to September reached 70.94%.

A serious AI visibility program therefore needs repeated observation. One monthly comparison is not enough to characterize the stability of a market's source ecosystem.

The supporting article How Stable Are AI Citation Sources Month to Month? A CMO Guide to Citation Persistence examines this temporal dimension in greater detail.

Cross-Platform Portability Was Low, but Not Equally Low for Every Platform Pair

Questions This Section Answers

1. Which AI platforms shared the most and least Top-100 citation sources across commercial markets?
2. Can a CMO assume that citation authority on one AI platform will transfer to another?
3. How different are the source ecosystems of ChatGPT, Gemini, Perplexity, Copilot, Google AI Overviews, and Google AI Mode?

The median market-level portability score was 22.70%, but the 15 platform pairs varied substantially.

The most important result is the separation between the top pair and everything else.

Google AI Overviews and Google AI Mode had a median overlap of 53.85%. No other pair exceeded 30.72%.

Even the two Google search surfaces were not close to identical. With two lists of exactly 100 domains, a 53.85% Jaccard overlap corresponds to 70 shared domains and 130 domains in the union. That still leaves 60 domain positions contributed by sources appearing on only one of the two lists.

The rest of the platform matrix is much more fragmented.

Gemini overlapped with the two Google search surfaces at 30.72%. Perplexity was nearly as close to each at 29.03%. ChatGPT's overlap with the Google search surfaces was 21.21%. Microsoft Copilot's overlap with them fell below 18%.

The dedicated Six AI Engines, Six Different Webs study expands this matrix, while the Google AI Overviews and AI Mode Are Siblings, Not Twins analysis focuses on the unusually close Google search-surface relationship and the surprising proximity of Perplexity to Gemini by this citation-overlap measure.

The Gap Appeared Across Every Major Macro Sector

Questions This Section Answers

1. Does the Persistence-Portability Gap appear across finance, health, legal, home, travel, consumer, and technology markets?
2. Which broad industries had the largest and smallest median gaps between citation persistence and portability?
3. Should a CMO use one global AI citation benchmark across every industry?

Every named macro sector had a positive median gap.

Finance & Insurance contains the sole negative market, Home Equity Loans, which is why it is the only named macro sector below 100% positive despite a strongly positive sector median.

The differences between sectors are meaningful, but they should not be overinterpreted. Market counts vary, exposure varies, and the taxonomy was not designed as a randomized sector experiment.

Still, the range is strategically useful. The median gap in Consumer & Lifestyle was about 13 points larger than in Legal & Compliance. A company benchmarking itself only against the overall 39.68-point median could miss important category structure.

The detailed sector view shows even wider differences.

This reinforces a key measurement principle: industry context matters even when the overall directional pattern is highly consistent.

The Most Prominent Sources Were Even More Persistent Relative to Their Portability

Questions This Section Answers

1. Does the Persistence-Portability Gap change when analysis focuses on the Top 25 or Top 50 AI citation sources instead of the Top 100?
2. Are the most prominent AI citation sources more durable over time than the broader source portfolio?
3. How sensitive is the 283-market result to the number of ranked domains included in each portfolio?

The main analysis uses exactly 100 domains in each eligible portfolio. The study repeated the calculation at Top 25 and Top 50 while retaining the same 283 markets and baseline cell-support requirement.

The gap becomes wider when attention narrows to the most prominent domains.

That does not mean Top-25 sources are universally better, more trustworthy, or more commercially valuable. It means the observed leading source portfolio was especially durable over time relative to its cross-platform transfer.

The finding is useful for CMOs because it suggests that the source environment has layers. A smaller core of highly prominent domains can be relatively stable, while the broader source set adds more turnover and more platform-specific variation.

The supporting analysis Top 25 vs Top 100 AI Citation Sources: Where Durable Authority Actually Lives examines this portfolio-depth effect in more detail.

The Result Survived Multiple Robustness Checks

Questions This Section Answers

1. Does the Persistence-Portability Gap remain when markets with taxonomy concerns or date proxies are excluded?
2. Does the 283-market result depend on incomplete platform or month coverage?
3. How robust is the median gap to weighting and alternative sample restrictions?

The benchmark was subjected to several descriptive sensitivity checks.

The result barely moves under these restrictions.

The no-proxy subset is especially important. Of 283 baseline markets, 183 had no ingestion-date proxy responses. Every one of those 183 markets had a positive gap, and the median gap was actually slightly higher than the baseline at 40.29 points.

That does not prove that removing proxy-dated responses individually from the full cohort would yield the same answer. The frozen aggregates do not support a full response-level reranking after excluding every proxy record. The legitimate statement is narrower: the market-level subset with no proxies remained fully positive.

The balanced-coverage subset also remained nearly identical. Among 246 markets with all three months and all six platforms represented in the marginal portfolios, 245 had positive gaps and the median was 39.21 points.

Taxonomy exclusions produced the same pattern. Removing 21 markets with current alias or unassigned-sector concerns left 262 markets, of which 261 remained positive, with a median gap of 39.88 points.

These checks make it difficult to attribute the headline finding to one obvious analytical choice such as portfolio depth, market weighting, taxonomy concerns, date proxies, or incomplete marginal coverage.

They do not solve the study's larger limitation: the data does not establish a certified fixed panel of identical prompts, model versions, countries, retrieval settings, and collection configurations.

Home Equity Loans Was the One Market That Reversed the Pattern

Questions This Section Answers

1. Which market was the only one where AI citation portability exceeded persistence?
2. Why should the Home Equity Loans result be preserved instead of treated as an error or outlier?
3. What does the sole negative-gap market reveal about using global AI citation benchmarks?

Home Equity Loans was the only qualifying market where portability exceeded persistence.

This is not a thinly supported comparison. The market had all three months and all six platforms represented, producing three month-pair comparisons and all 15 platform-pair comparisons.

Its unusual result came from very low temporal persistence. At 18.34%, Home Equity Loans had the lowest market-level persistence in the entire qualifying cohort.

The appropriate conclusion is not that this one market invalidates the other 282. Nor should it be removed merely because it is inconvenient.

It demonstrates that market-level exceptions are real and strategically important.

A CMO in home equity lending should not adopt the overall 63.93% persistence benchmark as a local expectation without inspecting the category. The benchmark provides context, not a substitute for market measurement.

The market also contained 600 ingestion-date proxy responses. Because the frozen aggregate package cannot rerank Home Equity Loans after removing those individual records, the exact role of those proxies is not estimable from the available evidence. The result should therefore be reported, not overexplained.

What This Means for CMOs

Questions This Section Answers

1. How should a CMO change AI search measurement if citation sources persist over time but do not transfer reliably across platforms?
2. Should marketing teams use one universal publisher target list for ChatGPT, Gemini, Perplexity, Copilot, and Google AI?
3. Which AI citation metrics belong in executive reporting alongside visibility and recommendation data?

The first implication is simple: do not confuse a durable source with a universal source.

A publisher that repeatedly appears in a market on ChatGPT can be strategically important even if its footprint is weak on Google AI Mode. A domain that matters across Google AI Overviews and AI Mode may still be relatively unimportant to Copilot.

For executive measurement, at least four dimensions should remain separate:

1. Citation depth: How prominent is the source inside the relevant market and platform?
2. Citation persistence: Does the source or source portfolio recur across time?
3. Citation portability: Does authority transfer across platforms?
4. Recommendation authority: Is the company, product, or service actually recommended?

The fourth dimension is deliberately separate. Being used as evidence is not the same thing as being recommended. The companion article AI Citation Authority vs AI Recommendation Authority: Why Being a Source Is Not the Same as Being Recommended develops that distinction.

For publisher outreach, the benchmark argues against one generic “AI authority” target list. Instead, build source maps at the intersection of market, platform, and time.

For budgeting, the data does not tell a CMO how much to spend on any publisher. Citation authority does not establish traffic, audience fit, media cost, brand lift, recommendation influence, or revenue. It tells the CMO where source authority is being observed and whether that authority appears durable or portable.

That is a much narrower claim, but it is also much more useful than pretending every frequently cited domain has the same strategic role.

What This Means for Publishers

Questions This Section Answers

1. How can a publisher distinguish durable AI citation authority from cross-platform citation breadth when talking to advertisers?
2. Can a publisher have commercially valuable AI citation authority even if that authority is concentrated on one platform?
3. What evidence should publishers show advertisers before making claims about AI visibility?

For publishers, the gap creates two separate commercial stories.

Durability asks whether a publisher remains part of a market's citation environment over time.

Transferability asks whether that authority appears across multiple AI platforms.

A publisher can be strong on one and weak on the other.

That matters because “AI visibility” by itself is too vague for an advertiser. A publisher should be able to say something more specific, such as:

• it appears repeatedly in a certain commercial market,
• it holds strong source positions on selected platforms,
• its authority persists over multiple collection periods,
• or its citation footprint transfers across several AI surfaces.

Those claims should be backed by market-specific evidence, cited URLs, observation windows, platform support, and appropriate disclosure.

The supporting article How Publishers Can Turn Persistent AI Citation Authority Into a Sponsorship and Distribution Asset focuses on how to make that commercial case without promising rankings or recommendation outcomes.

Why One Global GEO Strategy Is Unlikely to Be Enough

Questions This Section Answers

1. Can one GEO strategy reliably cover every AI platform and commercial market?
2. Why can a source strategy that works for one AI platform fail to transfer to another?
3. How should platform and industry differences change AI search optimization planning?

The benchmark does not say organizations need six completely separate marketing departments.

It does say that one source map is unlikely to represent the entire AI ecosystem.

The median portability of 22.70% is the strongest evidence for that conclusion. The platform-pair matrix reinforces it. Outside the Google AI Overviews and AI Mode pair, no median platform-pair overlap exceeded 30.72%.

That does not require six unrelated content strategies. Companies can still maintain:

• one factual source of truth,
• one brand positioning system,
• one content quality standard,
• one structured data framework,
• and one coordinated AI visibility program.

But the evidence supports platform-specific measurement and source acquisition inside that coordinated program.

The same logic applies by industry. Legal Services had a 30.35-point median gap. Business Software & Operations had 46.63 points. The market-level extremes were even wider.

This is why the broader research series includes There Is No Universal GEO Strategy and Can One GEO Strategy Work Across ChatGPT, Gemini, Perplexity, Copilot, Google AI Overviews, and Google AI Mode?.

Study Methodology

Questions This Section Answers

1. How were citation persistence and portability calculated across the 283-market AI citation benchmark?
2. What minimum data support was required before a month or platform portfolio could enter the comparison?
3. Which methodological limits should researchers consider before interpreting the Persistence-Portability Gap?

The primary benchmark covers completed observations from July, August, and September 2026 across six consumer AI surfaces:

• ChatGPT
• Gemini
• Perplexity
• Microsoft Copilot
• Google AI Overviews
• Google AI Mode

The strict qualifying cohort contains:

• 283 current tracked market IDs
• 876,605 retained responses
• 8,718,755 citation-array slots
• 126 unresolved slots
• 8,718,629 citation occurrences resolved to registrable domains
• 6,825,821 distinct response-domain appearances
• 308,807 distinct registrable domains

Ranking domains

Within each eligible market-month or market-platform cell, domains were ranked by the number of distinct retained responses in which the registrable domain appeared.

Repeated citations to the same domain inside one response counted once for portfolio ranking purposes.

Ties were broken alphabetically.

Eligibility rule

A comparison cell required at least:

• 100 retained responses
• 100 resolved registrable domains
• an actual ranked list of exactly 100 domains

Missing or thin cells were excluded. They were not assigned zero overlap.

Persistence

For persistence, the six consumer surfaces were pooled within each month. Eligible monthly Top-100 domain portfolios were compared using Jaccard similarity, defined as:

intersection / union

A market's persistence score is the median of its eligible month-pair overlaps.

Portability

For portability, completed months were pooled within each platform. Eligible platform-specific Top-100 portfolios were compared using the same Jaccard method.

A market's portability score is the median of its eligible platform-pair overlaps.

Gap

For each market:

Persistence-Portability Gap = market persistence - market portability

The headline 39.68-point median gap is the median of those 283 within-market differences.

It is not simply the difference between two independently summarized populations, although the median persistence minus median portability is numerically similar.

Overall summaries

Headline values are unweighted medians across the same qualifying markets unless otherwise labeled.

Descriptive bootstrap intervals use 10,000 market-level resamples with a fixed seed. Because markets can share domains, prompts, companies, and source ecosystems, these intervals should not be read as classical independent-sample population guarantees.

The evidence package was independently recalculated across all 283 markets and every eligible pair at Top 25, Top 50, and Top 100. It passed 20,232 deterministic QA checks with zero failures, subject to the documented source limitations.

Important Limitations

This is an observational analysis of stored AI citation samples.

It supports strong descriptive claims about the observed source portfolios. It does not support several tempting but stronger claims.

No certified identical-prompt panel

The historical collection does not establish an identical panel of prompts, prompt versions, models, countries, retrieval settings, and collection configurations for every month and platform.

The study therefore describes changing observed samples rather than a fixed-prompt causal experiment.

Date provenance is imperfect

104,071 responses, 11.87% of the strict cohort, use ingestion-date proxies.

The 183-market no-proxy subset remained fully positive, but a full response-level explicit-date-only reranking of all 283 markets is not available from the frozen aggregates.

Taxonomy is retrospective

The market and sector taxonomy was applied to historical observations. Twenty-one qualifying markets contain current alias or unassigned-sector concerns.

The conservative exclusion of those markets retained essentially the same result, but taxonomy should still be treated as part of the measurement system rather than an immutable natural classification.

Cross-job duplicates are candidates, not proven corrections

Exact fingerprint diagnostics identified candidate repeated observations across jobs. They may include genuine repeated observations. The study therefore preserved the baseline cohort rather than substituting a speculative deduplicated population.

Domains are not publishers

A registrable domain can represent a publisher, company, marketplace, community, government source, platform, or other entity.

Domain presence does not prove:

• publisher ownership,
• editorial independence,
• article quality,
• factual accuracy,
• traffic,
• recommendation influence,
• or commercial value.

Citations are not recommendations

The study measures source portfolios. It does not establish that a citation caused an AI platform to recommend a company, nor that citation authority will produce future rankings, traffic, conversions, or revenue.

The Larger Research Archive Is Broader, but It Is a Different Evidence Layer

The broader historical research archive reports 14,928,839 retained citation records across 1,616,287 retained responses.

That archive is valuable for source accounting, historical comparisons, and broader research development, but its response-level database was not available for complete independent replay in the flagship handoff.

For that reason, every primary conclusion in this article uses the stricter publication-grade cohort of 876,605 responses and 8,718,629 resolved-domain citation occurrences.

The two populations should not be added together or treated as interchangeable.

Conclusion

The main finding from the 283-market benchmark is not that AI citations are stable.

They are not.

The median market still changed meaningfully from month to month, and the July-September overlap was considerably lower than the August-September overlap.

The stronger finding is comparative:

AI citation-source portfolios were far more similar across completed months than across AI platforms.

That pattern appeared in 282 of 283 markets.

The median gap was 39.68 percentage points. The result remained near 40 points under response weighting, taxonomy exclusions, no-proxy market restrictions, and balanced three-month/six-platform coverage. Narrowing the portfolios to the Top 25 sources widened the gap to 46.99 points.

The market examples show why this matters. Long-Term Care Insurance combined 85.19% persistence with 23.46% portability. Life Insurance Companies combined 72.41% persistence with 14.94% portability. Stairlifts produced a 55.04-point gap across 4,600 retained responses. Yet Home Equity Loans reversed the pattern entirely.

The platform matrix shows the same fragmentation. Google AI Overviews and AI Mode were much more similar to one another than any other pair, but even they shared only a 53.85% median Jaccard overlap. Most other platform pairs fell between roughly 16% and 31%.

For CMOs, publishers, agencies, and researchers, that leads to a more precise model of AI authority:

Source authority has to be measured across time, platform, market, and depth.

A durable source is not automatically portable. A portable source is not automatically prominent. A cited source is not automatically a recommended company. And a global benchmark is not a substitute for a market-specific source map.

The next generation of AI search measurement should preserve those distinctions rather than collapsing them into one visibility score.

Related LLM Authority Index Research

• The Persistence-Portability Gap: Why AI Citation Authority Persists Over Time but Fragments Across Platforms introduces the conceptual framework.
• How Stable Are AI Citation Sources Month to Month? A CMO Guide to Citation Persistence focuses on temporal source stability.
• Persistence vs Portability: Why a Source That Wins on ChatGPT May Not Transfer to Gemini or Google AI examines why durable source authority does not automatically transfer across platforms.
• Top 25 vs Top 100 AI Citation Sources: Where Durable Authority Actually Lives examines portfolio-depth sensitivity.
• How Publishers Can Turn Persistent AI Citation Authority Into a Sponsorship and Distribution Asset translates persistence into publisher-side commercial measurement.
• AI Search Has a Tiny Universal Core and a Massive Specialist Long Tail examines how domain authority fragments across markets.
• Six AI Engines, Six Different Webs provides the full six-platform citation-overlap comparison.
• There Is No Universal GEO Strategy examines sector and industry variation.
• The 2026 AI Commercial Citation Authority Index separates domain breadth from depth across the benchmark.
• AI Citation Authority vs AI Recommendation Authority distinguishes source authority from company recommendation behavior.
• The 2026 AI Citation Authority Study documents the earlier 53-industry high-stakes consumer benchmark.
• Do AI Platforms Cite the Same Websites for High-Stakes Consumer Decisions? provides earlier cross-platform evidence.
• How Stable Are AI Citations for High-Stakes Consumer Decisions? examines repeated-query citation drift in the earlier research program.

About LLM Authority Index

LLM Authority Index measures how AI systems cite sources, surface publishers, and structure authority across commercial markets. Its research program separates citation authority, source persistence, platform portability, market breadth, source depth, recommendation behavior, and other dimensions that are often collapsed into a single AI visibility score.