Does Cross-Platform AI Visibility Matter? Measuring Recommendation Portability and Platform Concentration Risk

Research status: Exploratory longitudinal research. Cross-platform AI recommendation breadth has not been validated as a predictor of revenue, earnings, analyst revisions, valuation, or stock returns.

Initial observation window: July through September 2026

Initial public-company panel: 25 mapped public parents

Broader company/entity momentum table: 406 companies and entities

AI platform families: ChatGPT, Gemini, Google AI Mode, Google AI Overviews, Microsoft Copilot, and Perplexity

Current methodology version: V0

Answer Capsule

Cross-platform AI visibility likely matters because a company that gains or loses recommendation coverage across several independent AI systems presents a different measurement pattern from a company whose movement is concentrated in one platform.

The initial LLM Authority Index public-company panel shows exactly why platform breadth should be measured separately from aggregate recommendation change.

Among the 24 public-company rows with a nonzero aggregate recommendation-coverage change, 18 moved in the same direction on at least four of the six measured platform families, 9 moved in the same direction on at least five platforms, and 3 moved in the same direction on all six platforms.

Those three all-platform directional cases were Goldman Sachs / Marcus, Bank of America, and Happen / LendingClub, each of which showed declining recommendation coverage on all six platform families during the matched observation window.

But aggregate direction can also hide fragmentation. Lincoln Financial had an aggregate recommendation-coverage change of -6.5 percentage points, yet only two platforms moved downward while four moved upward. Trupanion declined by -5.4 points overall, but the six platforms split evenly, three negative and three positive. Axos Financial gained +19.7 points overall and improved on four platforms, while Perplexity moved sharply in the opposite direction and Gemini was flat.

These examples support a basic measurement rule:

Aggregate AI recommendation momentum and cross-platform portability are related, but they are not the same thing.

The current V0 framework therefore treats platform breadth as a separate signal-quality dimension. It does not assume that agreement across more platforms makes a company financially stronger or weaker. The forward-looking question is whether persistent, multi-platform recommendation movement contains more useful commercial information than platform-specific movement.

That hypothesis will be tested through the prospective validation framework described in Can AI Search Visibility Predict Revenue Growth?. The core AI-side measurement rules are documented in How We Measure AI Commercial Momentum, and the current company observations are preserved in the AI Investor Signal Tracker.

Key Findings From the Initial Public-Company Panel

The three six-of-six directional cases were:

• Goldman Sachs / Marcus: aggregate recommendation coverage -11.3 percentage points, with all six platform families declining.
• Bank of America: aggregate recommendation coverage -8.6 points, with all six platform families declining.
• Happen / LendingClub: aggregate recommendation coverage -8.2 points, with all six platform families declining.

The strongest positive V0 examples were not fully portable:

• MetLife: +11.3 points overall, with five platforms improving and Perplexity declining.
• Axos Financial: +19.7 points overall, with four platforms improving, Gemini flat, and Perplexity declining.

This is important because the size of an aggregate change and the breadth of platform agreement answer different questions.

A large aggregate gain can be concentrated in a subset of systems. A smaller aggregate change can be directionally consistent across nearly every platform. Both patterns may matter, but they should not be interpreted as interchangeable.

Questions This Section Answers

• What is recommendation portability?
• What is platform concentration risk in AI search?
• Why should investors care whether several AI systems move together?

What Is Recommendation Portability?

Recommendation portability describes how consistently a company’s recommendation movement transfers across different AI platform families.

In the current investor framework, portability is not a statement about citations or source authority. It is specifically about company recommendation direction.

For example, if a company’s matched recommendation coverage rises from July to September on ChatGPT, Gemini, Google AI Mode, Google AI Overviews, Microsoft Copilot, and Perplexity, the direction is highly portable across platforms.

If the aggregate recommendation rate rises but only two systems improve while four decline, the aggregate movement is not broadly portable.

This distinction matters because every AI platform operates through a different combination of model architecture, retrieval systems, search indexes, source ecosystems, prompt handling, answer construction, ranking behavior, and product design.

Current third-party research reinforces that AI systems should not be treated as one homogeneous discovery channel. Ahrefs reports that the overlap among the top cited domains on Google AI Overviews, ChatGPT, and Perplexity can be limited, and its platform-comparison work emphasizes that different AI systems rank and cite content differently. Ahrefs has also reported substantial differences between Google AI Mode and AI Overviews even when the resulting answers are semantically similar. See Ahrefs, How Ranking in Google AI Overviews, ChatGPT, and Perplexity Are Different.

BrightEdge has separately documented cases where ChatGPT and Google AI Overviews cite different types of assets from the same social platforms, further illustrating that AI discovery behavior can vary by engine. See BrightEdge, Profiles Over Posts.

Those citation studies do not prove anything about public-company recommendation momentum or investment performance. They support a narrower premise: different AI platforms can produce materially different visibility ecosystems.

That makes portability a measurement variable worth preserving rather than averaging away.

What Is Platform Concentration Risk?

Platform concentration risk is the possibility that a company’s apparent AI momentum depends disproportionately on one platform or a small subset of platforms.

Suppose a company’s aggregate recommendation coverage rises by 10 percentage points, but almost all of that increase comes from one platform while the other five systems are flat or declining.

That result may still be commercially relevant. If the platform is large, growing quickly, and important to the company’s customer acquisition funnel, a concentrated signal could matter.

But it is a different research object from a 10-point increase that appears independently across five or six systems.

The concentrated case raises additional questions:

• Is the movement caused by a platform-specific model or retrieval update?
• Does one platform disproportionately expose the company to a certain prompt cluster?
• Is the change likely to persist after the platform changes models, ranking rules, or source preferences?
• Does the platform have meaningful consumer reach in the measured category?
• Does the same commercial signal appear in other behavioral data?

For investor research, platform concentration therefore acts more like a risk characteristic of the AI measurement than a financial-risk conclusion about the company.

Platform Breadth Is Already Built Into the V0 Candidate Rules

The current V0 methodology does not classify a company as a positive or negative AI divergence candidate based only on the size of the aggregate recommendation change.

A positive candidate requires:

1. recommendation coverage change of at least +5 percentage points;
2. an exploratory 95% interval with a lower bound above zero; and
3. at least four platform families improving.

A negative candidate requires:

1. recommendation coverage change of at least -5 percentage points;
2. an interval with an upper bound below zero; and
3. at least four platform families worsening.

This platform-breadth requirement is deliberate.

A large change concentrated in one or two engines can still be interesting, but it does not qualify for the same watch category as a change supported by most of the measured platforms.

For the full V0 construction rules, see How We Measure AI Commercial Momentum.

The Initial Portability Spectrum

The V0 public-company panel contains several different cross-platform patterns.

Fully portable directional movement

Three companies moved in the same direction on all six platform families.

These are strong examples of directional portability inside the initial measurement window.

They are not evidence that future revenue, earnings, or stock performance will decline. The only conclusion supported by the table is that recommendation coverage weakened across every measured AI platform family for these tracked entities during the matched comparison.

The individual company analyses are linked through the AI Investor Signal Tracker.

Broad but not universal movement

MetLife is the clearest positive example.

MetLife’s aggregate recommendation coverage increased by 11.3 percentage points. Five platform families improved:

• ChatGPT: +23.08 points
• Gemini: +15.38
• Google AI Mode: +8.51
• Google AI Overviews: +9.84
• Microsoft Copilot: +26.32

Perplexity declined by -5.41 points.

The result is broad, but not universal.

That distinction is useful because it gives us a concrete future test. If later commercial outcomes improve, does a five-of-six pattern contain similar information to a six-of-six pattern? Does the dissenting platform matter? Are some platforms more commercially informative than others?

Those questions cannot be answered from the current three-month AI panel alone.

Large aggregate movement with meaningful disagreement

Axos Financial had the strongest positive aggregate change in the V0 public panel at +19.7 percentage points.

The platform pattern was:

• ChatGPT: +14.29
• Gemini: 0.00
• Google AI Mode: +43.14
• Google AI Overviews: +29.03
• Microsoft Copilot: +18.18
• Perplexity: -19.23

Four platforms improved, one was flat, and one moved sharply in the opposite direction.

This is precisely why the aggregate result and the portability result should remain separate.

The aggregate signal says Axos Financial’s mapped entities gained recommendation coverage substantially in the matched panel. The portability layer says the gain was broad enough to qualify under V0, but not universal, and Perplexity moved materially against the aggregate direction.

The company-level article is Axos Financial AI Search Visibility.

Questions This Section Answers

• Can an aggregate AI signal hide platform disagreement?
• What does a split platform pattern look like?
• Why might platform disagreement be analytically useful rather than just noise?

When the Aggregate Number Hides a Split Market

Two companies illustrate the danger of reading only the aggregate recommendation change.

Lincoln Financial

Lincoln Financial’s aggregate recommendation coverage declined by 6.5 percentage points.

But the platform deltas were:

• ChatGPT: +62.50
• Gemini: +9.09
• Google AI Mode: +5.56
• Google AI Overviews: -17.19
• Microsoft Copilot: +9.09
• Perplexity: -45.45

Only two platform families moved in the same negative direction as the aggregate result. Four moved positively.

The V0 panel therefore classifies Lincoln Financial as mixed or neutral rather than as a negative AI divergence candidate, despite the size of the aggregate decline.

That is an important design choice.

The aggregate decline may reflect weighting, prompt availability, platform-level sample structure, or larger negative changes on the platforms that moved down. But the cross-platform evidence does not support describing the movement as broadly negative across AI systems.

Trupanion

Trupanion’s aggregate recommendation coverage declined by 5.4 percentage points.

Its platform pattern split three against three:

• ChatGPT: +46.15
• Gemini: +3.85
• Google AI Mode: -12.77
• Google AI Overviews: -4.92
• Microsoft Copilot: +5.26
• Perplexity: -32.43

The aggregate result is negative, but the platform ecosystem is fragmented.

This is not necessarily a measurement failure. Fragmentation can itself be informative.

Different systems may be responding to different source markets, different retrieval indexes, different answer policies, different company associations, or different slices of the prompt universe.

The research question is whether that disagreement is stable enough to characterize and whether certain forms of platform disagreement have predictable downstream implications.

At present, we do not know.

Recommendation Portability vs. the Persistence-Portability Gap

LLM Authority Index already uses the term Persistence-Portability Gap in its citation-source research.

That framework was developed from a different measurement layer: citation-source authority, not company recommendation behavior.

In the high-stakes consumer citation panel, 88 domains remained in the overall monthly Top 100 across July, August, and September 2026, while only 15 domains appeared in the Top 100 across all six AI platform families. Using like-for-like median pairwise Jaccard similarity, monthly source persistence measured 85.2% while cross-platform portability measured 36.1%, producing a study-specific 49.1 percentage-point Persistence-Portability Gap.

See The Persistence-Portability Gap: Why AI Citation Authority Persists Over Time but Fragments Across Platforms.

The investor-signal research asks whether a related structural principle appears at the company recommendation layer.

The analogy is useful, but the metrics must not be merged.

Citation-source portability

Citation-source portability asks:

Does a publisher or source domain that is important on one AI platform also remain important on other AI platforms?

Recommendation portability

Recommendation portability asks:

Does a company’s direction of recommendation change transfer across different AI platform families?

These are different questions with different denominators.

A publisher can be widely portable as a source without any particular company being widely portable as a recommendation. A company can gain recommendations across multiple platforms even if the supporting source ecosystems are different.

For investor research, recommendation portability is the more direct variable because it measures the company-selection layer. Citation portability remains useful for explaining why recommendation behavior may differ across systems.

A Proposed Recommendation Portability Framework

The current V0 panel uses a simple and transparent platform-breadth count.

For a company with a positive aggregate recommendation change:

Positive portability count = number of platform families with positive matched recommendation change

For a company with a negative aggregate recommendation change:

Negative portability count = number of platform families with negative matched recommendation change

A simple directional portability ratio can therefore be expressed as:

Directional Portability Ratio = same-direction platforms / observed platform families

Examples from the initial panel:

• Goldman Sachs / Marcus: 6/6 = 100%
• Bank of America: 6/6 = 100%
• Happen / LendingClub: 6/6 = 100%
• MetLife: 5/6 = 83.3%
• Axos Financial: 4/6 = 66.7%, with one flat platform
• Trupanion: 3/6 = 50.0%
• Lincoln Financial: 2/6 = 33.3%

This ratio is intentionally simple.

It should not yet be converted into a single investment score because platform size, commercial influence, prompt coverage, signal magnitude, and variance differ across systems.

A 10-point movement on one platform may not be economically equivalent to a 10-point movement on another platform.

Future versions may therefore test additional portability measures, such as:

• magnitude-weighted platform agreement;
• platform-specific historical predictive weights;
• sign consistency over consecutive months;
• prompt-cluster-specific platform agreement;
• platform dispersion or variance;
• entropy or concentration measures;
• and persistence-adjusted portability.

Those extensions should only be adopted if they improve out-of-sample validation rather than because they make the historical fit look cleaner.

Platform Concentration Is Not Automatically Bad

A common mistake would be to assume that multi-platform movement is always superior and concentrated movement is always irrelevant.

That is not necessarily true.

A platform-specific signal could matter if:

• that platform has disproportionate usage in the relevant buyer journey;
• the platform sends measurable referral traffic or influences branded search;
• the platform specializes in the category;
• the company’s target customer segment over-indexes on that platform;
• or the platform-specific change consistently precedes commercial outcomes.

For example, if future testing showed that one platform’s recommendation changes lead branded search or revenue revisions in a specific category, then a concentrated signal on that platform could be economically important even without broad cross-platform agreement.

The correct empirical question is therefore not:

Is broad portability always better?

It is:

Does platform breadth improve the forward information content of AI recommendation momentum after controlling for signal magnitude, persistence, company exposure, and ordinary financial variables?

That question is part of the validation program described in Can AI Search Visibility Predict Revenue Growth?.

Why Platform Agreement Might Improve Signal Quality

There are several plausible reasons why cross-platform agreement could matter.

1. It can reduce dependence on one model update

AI products change quickly. A recommendation gain caused by one retrieval-system adjustment may reverse when that platform changes models or source policies.

A movement observed independently across several systems may be less likely to be a single-platform artifact.

That is a hypothesis, not a proven rule.

2. It can indicate broader information-environment change

If several AI systems begin surfacing the same company more frequently, the underlying change may reflect something broader than one platform’s internal ranking behavior.

Possible drivers include:

• stronger third-party coverage;
• more consistent product positioning;
• changing review consensus;
• increased publisher authority;
• improved company-owned information;
• market events;
• greater brand prominence;
• or broader changes in the web evidence available to retrieval systems.

The current dataset does not identify causality among those mechanisms.

3. It can reduce the chance that an aggregate score is being driven by one extreme platform

A cell-weighted aggregate can move even when platform directions disagree.

Breadth provides a second dimension that helps distinguish broad movement from concentrated movement.

4. It may improve robustness across user behavior

Consumers do not all use the same AI platform.

If AI-mediated discovery becomes economically important, commercial exposure may be distributed across multiple systems. A company visible across several platforms could therefore have a more diversified AI discovery footprint.

Again, this does not prove a revenue effect. It identifies a measurable exposure characteristic.

Why Platform Disagreement Could Also Be Valuable

Cross-platform disagreement is not merely a nuisance variable.

It can reveal structure that an aggregate score hides.

A split signal can help identify:

• platform-specific strengths and weaknesses;
• source-market fragmentation;
• category differences;
• prompt clusters that produce divergent recommendation sets;
• changes that may be temporary or platform-specific;
• and opportunities to test which systems, if any, have greater downstream commercial relevance.

For investors, that means the disagreement itself can become a research feature.

If one platform repeatedly leads the others before commercial outcomes change, that platform may deserve more predictive weight later.

If platform disagreement produces no stable downstream relationship, breadth may become a useful confidence filter rather than a predictive variable.

Either result would be informative.

Questions This Section Answers

• How will LLM Authority Index test whether platform breadth adds predictive value?
• What would support the portability hypothesis?
• What would count against it?

How We Will Test Whether Cross-Platform Breadth Matters

The formal validation should compare models rather than rely on narrative interpretation.

A simplified hierarchy is:

Model A: baseline without AI variables

Potential baseline inputs include information already available at the signal date, such as:

• prior revenue growth;
• prior analyst revisions;
• valuation variables;
• sector and macro controls;
• branded search or web traffic where available;
• company size;
• seasonality;
• and other conventional alternative-data inputs available before the outcome.

Model B: baseline plus aggregate AI recommendation momentum

This tests whether the aggregate AI signal adds incremental information.

Model C: baseline plus aggregate AI momentum and platform breadth

This adds variables such as:

• number of platforms moving in the aggregate direction;
• directional portability ratio;
• number of opposing platforms;
• number of stable platforms;
• and platform dispersion.

If Model C consistently improves forward performance over Model B, platform breadth may contain incremental information beyond aggregate AI momentum.

Model D: platform-specific AI variables

This model would allow each platform to enter separately.

If one or two platforms consistently dominate predictive performance, a platform-specific model may outperform a simple breadth measure.

This possibility is important because the research should not force a multi-platform theory if the evidence later shows that certain systems are much more commercially informative than others.

The full prospective validation architecture is detailed in Can AI Search Visibility Predict Revenue Growth?.

Predeclared Portability Hypotheses

The following hypotheses can be evaluated prospectively.

H1: Multi-platform movement will be more informative than isolated movement

If true, recommendation changes supported by four, five, or six platforms should produce stronger forward relationships with commercial outcomes than changes driven by one or two platforms.

H2: Persistent multi-platform movement will be more informative than one-month agreement

A company that moves in the same direction across several platforms for multiple consecutive months may produce a stronger signal than a company that briefly reaches broad agreement once.

H3: Platform breadth may matter more in some sectors than others

Recommendation-sensitive consumer categories may benefit more from cross-platform breadth than industries where AI discovery has little connection to purchasing behavior.

H4: Some platforms may contain distinct information even when portability is low

A low-portability signal could still matter if one platform consistently leads later outcomes.

H5: Citation-source portability and company-recommendation portability may be related but not identical

Future work can test whether companies gain broader recommendation coverage when their supporting source ecosystem is also more portable across AI systems.

Each hypothesis may prove false.

What Would Count Against the Portability Hypothesis?

Cross-platform breadth would be less useful than expected if future testing finds that:

• platform breadth does not improve prediction beyond aggregate recommendation change;
• six-of-six and two-of-six signals have similar forward information value after controlling for magnitude;
• platform agreement is unstable from month to month;
• one dominant platform explains nearly all of the downstream relationship;
• breadth mostly reflects shared source exposure rather than independent evidence;
• results disappear out of sample;
• or platform breadth adds complexity without improving forecasting accuracy or decision quality.

If those patterns emerge, the framework should be narrowed.

The broader project has already published a separate falsification framework in What Would Prove the AI Commercial Momentum Hypothesis Wrong?.

Sector Research Will Make Portability More Interpretable

The initial public-company panel mixes banks, insurers, fintech companies, lenders, healthcare companies, brokerages, and crypto-related businesses.

Cross-platform behavior may differ substantially by sector.

The next stage of the research therefore breaks the panel into dedicated sector studies:

• Bank Stocks and AI Search
• Insurance Stocks and AI Search
• Fintech, Brokerage and Crypto Stocks
• Mortgage and Lending Stocks in AI Search
• Healthcare Stocks and AI Search

Those studies should make it easier to distinguish sector structure from company-specific movement.

Investor Interpretation Framework

What Cross-Platform Visibility Does Not Mean

Cross-platform recommendation breadth does not establish:

• causal influence on purchases;
• market-share growth;
• revenue acceleration;
• revenue deterioration;
• analyst estimate revisions;
• earnings surprise;
• valuation mispricing;
• future stock returns;
• or the quality of a company’s products or services.

It measures one narrower fact: how broadly a direction of AI recommendation change appears across the measured platform families.

A company can lose recommendation coverage on all six systems and still grow revenue for reasons unrelated to consumer AI discovery.

A company can gain recommendation coverage on five systems and still disappoint investors because expectations were already high or because the measured brand represents only a small part of the public parent.

The AI Recommendations vs. Mentions vs. Citations article explains why platform breadth should also remain separate from citation, mention, rank, and sentiment measurements.

Methodology Notes

The V0 platform-breadth analysis uses the same matched-panel foundation as the broader investor-signal methodology.

For each mapped company or public parent:

1. The same normalized prompt and platform family must be available in the base month and September 2026.
2. Explicit extraction failures are excluded rather than treated as zero visibility.
3. Response-identical cross-vertical duplicate exports are collapsed in the primary cleaned panel.
4. Relevant brands and entity variants can be rolled to a public parent under governed mappings.
5. Recommendation coverage is calculated across eligible matched prompt-platform cells.
6. Each platform family receives a matched recommendation-change value.
7. Platform direction is classified as improving, worsening, or stable according to the sign of that matched change.
8. Aggregate recommendation change and platform breadth are retained as separate variables.

July 2026 is the base month when available. Travelers uses August because July is unavailable in its matched panel. Labcorp has five observed platform families in the current public row rather than six.

The current V0 confidence rules already incorporate platform count. High confidence requires all six platform families in addition to sample-size, interval, and sensitivity requirements.

Limitations

This cross-platform analysis has several important limitations.

1. Six platforms are not the entire AI ecosystem

The panel covers six major platform families, but not every AI assistant, model, embedded agent, shopping interface, vertical assistant, or enterprise search tool.

2. Platform usage is not weighted

The current breadth count does not weight platforms by active users, category-specific adoption, referral traffic, commercial influence, or revenue contribution.

3. Direction can hide magnitude differences

A +1-point platform move and a +30-point platform move both count as positive direction in a simple breadth statistic.

4. Aggregate and platform-level denominators may differ in practical influence

The aggregate point estimate is cell-weighted. Platform-level changes can be based on different numbers of matched prompt cells.

5. Platform behavior can change rapidly

Model versions, retrieval systems, search indexes, product interfaces, citation behavior, and recommendation policies can change after the observation period.

6. Prompt populations are research panels

The prompts are commercially relevant and unbranded, but they are not weighted by actual query volume, transaction value, customer lifetime value, or market size.

7. Brand-to-parent exposure can differ

Some tracked entities are products, divisions, subsidiaries, or legacy brands rather than the entire public company.

8. Portability has not been financially validated

The current results show platform agreement patterns. They do not show that broader agreement predicts stronger commercial or market outcomes.

What We Will Test Next

The next longitudinal stages should test whether platform breadth changes the usefulness of the AI signal.

The most important comparisons are:

1. aggregate AI momentum alone vs. aggregate momentum plus platform breadth;
2. high-portability signals vs. low-portability signals;
3. persistent multi-platform signals vs. one-month multi-platform signals;
4. platform-specific variables vs. an equal-weight breadth statistic;
5. sector-specific portability effects;
6. recommendation portability vs. citation-source portability;
7. platform breadth vs. future branded search and web/app engagement;
8. platform breadth vs. analyst revenue revisions and reported revenue outcomes;
9. platform breadth vs. sector-relative returns only after earlier commercial validation.

The formal backtesting architecture will be developed further in How Investors Could Backtest AI Search Signals Against Revenue, Analyst Estimates and Stock Performance.

Related LLM Authority Index Research

• Can AI Search Signal Future Revenue Growth? The AI Commercial Momentum Hypothesis
• How We Measure AI Commercial Momentum
• AI Investor Signal Tracker
• AI Recommendations vs. Mentions vs. Citations
• Can AI Search Visibility Predict Revenue Growth?
• The Persistence-Portability Gap
• Bank Stocks and AI Search
• Insurance Stocks and AI Search
• Fintech, Brokerage and Crypto Stocks
• Mortgage and Lending Stocks in AI Search
• Healthcare Stocks and AI Search

External Measurement Context

The following sources provide context for why AI platforms should be measured separately. They do not validate the financial hypotheses in this article.

1. Ahrefs. How Ranking in Google AI Overviews, ChatGPT, and Perplexity Are Different. https://ahrefs.com/academy/aeo-course/lesson-1-2
2. Ahrefs. Only 12% of AI Cited URLs Rank in Google's Top 10 for the Original Prompt. https://ahrefs.com/blog/ai-search-overlap/
3. BrightEdge. Profiles Over Posts: How ChatGPT and Google AI Overviews Cite Social Platforms Differently. https://help.brightedge.com/resources/weekly-ai-search-insights/profiles-over-posts-chatgpt-google-ai-overviews

Research Status Note

This article documents an exploratory measurement framework, not an investment recommendation.

The initial evidence shows that cross-platform agreement varies substantially by company and that aggregate AI recommendation movement can hide important platform disagreement. Whether platform breadth improves the prediction of later commercial or financial outcomes remains an open empirical question.

The correct next step is not to assume that six-platform agreement is financially superior. It is to test that claim prospectively.