AI Recommendation Share vs. Market Share: Could the Gap Reveal Emerging Commercial Strength or Weakness?

Research status: Prospective exploratory framework. AI recommendation share has not been validated as a predictor of market-share change, revenue growth, analyst revisions, valuation, or stock returns.

Initial AI 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

The gap between a company's AI recommendation share and its real-world market share could become a useful commercial research signal, but the relationship has not yet been validated.

The core idea is simple. If a company wins a materially larger share of unbranded AI recommendations than its current share of real-world demand, that could indicate that AI systems are exposing the company to a larger future consideration set than its present market position would suggest. If the company wins a materially smaller share of AI recommendations than its market share, the reverse could be true.

But that gap is not automatically evidence of emerging growth or decline.

AI systems may overrepresent specialist brands, premium brands, digitally visible companies, brands with strong third-party coverage, or companies that fit common prompt language. Real-world market share may reflect installed base, distribution, contracts, geography, pricing, enterprise relationships, regulation, legacy customers, offline channels, or business segments that consumer-facing AI prompts do not observe.

The research question is therefore not:

Does AI recommendation share equal market share?

It is:

Does the gap between AI recommendation share and market share contain forward-looking information about future commercial share, revenue growth, customer acquisition, or related outcomes?

That question must be tested prospectively.

The underlying theory is defined in The AI Commercial Momentum Hypothesis. The AI-side measurement system is documented in How We Measure AI Commercial Momentum. The live company observations are preserved in the AI Investor Signal Tracker.

The Core Research Concept

Questions This Section Answers

• What is AI recommendation share?
• How is it different from market share?
• What does a gap between the two actually mean?

AI recommendation share is the share of eligible AI recommendation opportunities captured by a company relative to competitors in a defined category, prompt universe, platform set, and observation period.

Market share is the company's share of actual economic activity in the corresponding real-world market. Depending on the sector, that might mean revenue, unit sales, customers, deposits, insurance premiums, loan originations, assets, subscribers, active users, transactions, or another economically meaningful denominator.

These are not the same measurement.

A recommendation share measures AI-mediated competitive selection.

A market share measures real-world economic participation.

The proposed research gap is:

AI Recommendation Share Gap = AI recommendation share - comparable real-world market share

A positive gap means the company captures a larger share of AI recommendations than its current market position would suggest.

A negative gap means the company captures a smaller share of AI recommendations than its current market position would suggest.

The gap is descriptive until future outcomes are tested.

It does not prove that the company's market share will converge toward its AI recommendation share.

Why the Gap Could Matter

The commercial logic is based on consideration.

When a user asks an unbranded question such as "What are the best online banks for high-yield savings?" or "Which pet insurance companies are best for older dogs?", an AI system can compress a large competitive market into a small recommendation set.

That can create a possible pathway:

Unbranded commercial prompt
-> AI recommendation set
-> brand enters consumer consideration
-> branded search or direct navigation
-> site, app, retailer, broker, or branch interaction
-> customer acquisition or purchase
-> future commercial share

The revenue-growth validation study is designed to test whether the early parts of that chain actually contain forward information.

If AI recommendation share proves unrelated to later commercial outcomes, then the gap may still be useful for AI-search strategy, but it should not be treated as a financial or investor signal.

What Existing Research Suggests

The idea that AI recommendation share can differ from real-world market share is already observable.

A 2026 working paper by Alejandro Medina Sandin compared brand recommendations from three frontier LLMs with real-world market shares across sixteen U.S. consumer categories. The study found that LLM recommendation concentration did not simply reproduce market concentration. In fragmented categories, AI recommendations tended to concentrate on fewer brands than the market. In highly concentrated categories, the models distributed recommendations more broadly than actual market shares would imply. The study also found a tilt toward specialist and premium brands among top recommendations. See AI Meets Antitrust: How Large Language Models Reshape Market Concentration.

That finding is important because it shows that AI recommendations can create a competitive representation of a market that differs from the market's current economic structure.

It does not show that the AI distribution predicts where market share moves next.

AIVO's revised LLM Equity Valuation framework provides another relevant precedent. Following a September 23, 2026 methodology correction, AIVO defined Organic Win Rate as the share of unprompted category conversations in which an assistant makes a brand its final recommendation. Its revised AI Share Index divides Organic Win Rate by the brand's market share. A ratio of 1.0 represents parity between AI recommendation share and market share. See AIVO's methodology correction.

AIVO also states an important limitation: the assumption that AI-influenced purchases distribute in proportion to final recommendation share has not been validated. Its revised framework specifically calls for longitudinal testing of whether changes in recommendation share precede changes in AI-referred revenue.

That limitation is central to the LLM Authority Index approach as well.

We do not currently assume that AI recommendation share becomes economic market share.

We want to test whether the direction and magnitude of the gap contain useful information about what happens later.

AI Recommendation Share Is Not the Same as Recommendation Coverage

Questions This Section Answers

• How is recommendation share different from recommendation coverage?
• Why does the investor research need both?
• Can a company improve coverage without gaining share?

The current V0 AI Investor Signal Tracker primarily reports recommendation coverage.

Recommendation coverage asks:

Across all eligible matched prompt-platform cells, how often was this company recommended?

Recommendation share asks a different question:

Among the recommendations allocated to companies in the competitive category, what portion went to this company?

Those metrics can move differently.

A company can increase recommendation coverage while competitors increase even faster, producing flat or declining competitive recommendation share.

A company can maintain similar coverage while competitors disappear, increasing its share of the remaining recommendation set.

A company can also dominate a narrow prompt cluster while holding modest recommendation share across the entire sector.

For investor research, recommendation coverage is useful for measuring a company's absolute AI consideration frequency. Recommendation share is useful for measuring its competitive position within a defined category.

Neither should be substituted for real-world market share.

The broader distinction between AI measurement types is explained in AI Recommendations vs. Mentions vs. Citations.

A Better Way to Think About the Gap

The recommendation-share gap should initially be treated as a competitive discrepancy, not a growth forecast.

There are four basic states:

None of these states is a stock rating.

A positive gap is not evidence that a company should be bought.

A negative gap is not evidence that a company should be sold.

The gap simply identifies cases where AI systems allocate competitive attention differently from the real economy.

Why Positive AI Recommendation Share Could Be Interesting

A positive recommendation-share gap could arise for several reasons.

1. The brand may be entering consideration faster than its current scale reflects

A smaller company can appear frequently in AI recommendations because its features fit the way consumers phrase unbranded questions.

If that exposure later produces increased branded search, traffic, account openings, sales, or customer acquisition, the gap may have been an early commercial signal.

2. The brand may have unusually strong digital evidence

AI systems may be finding strong reviews, comparisons, product documentation, category authority, expert commentary, or third-party support that makes a company easy to recommend.

That could create an AI advantage before the company has fully converted that advantage into market share.

3. The brand may fit high-intent niches particularly well

A company may not lead the overall category but can dominate important prompt clusters such as:

• best for budget-conscious buyers;
• best for a specific age group;
• best for a specialized product need;
• best for self-service users;
• best for premium buyers;
• best for particular geographies or use cases.

A broad market-share statistic can obscure those pockets of strength.

4. AI may be amplifying emerging specialist brands

The 2026 market-concentration study found evidence that top LLM recommendations can tilt toward specialist and premium brands rather than simply reproducing category leaders.

That could create a genuine discovery advantage.

It could also create a model bias unrelated to future economic performance.

Only longitudinal testing can distinguish those possibilities.

Why Negative AI Recommendation Share Could Be Interesting

A negative gap can also have several explanations.

1. The company may have strong legacy market share but weak AI-mediated consideration

Large incumbents can hold meaningful real-world share because of distribution, contracts, branch networks, installed base, brand memory, existing customer relationships, or offline channels.

AI systems may not allocate recommendation share in proportion to those advantages.

If AI-mediated discovery becomes economically important, that discrepancy could eventually matter.

2. The category may not be well represented by consumer AI prompts

A company can have large market share in enterprise, employer, wholesale, institutional, brokered, or regulated channels that are not captured well by direct-to-consumer prompt research.

In that case, a negative recommendation-share gap may simply reveal that the measurement universe is incomplete.

3. The AI systems may be stale or structurally biased

AI recommendations can reflect outdated facts, uneven retrieval, language bias, publisher availability, prompt framing, or source concentration.

A negative gap may therefore represent an AI-system limitation rather than commercial weakness.

4. The measured brand may be only one part of a diversified public company

The current V0 panel includes cases where the measured exposure comes from brands or products such as UFB Direct, Marcus, Coinbase Wallet, CVS Pharmacy, Labcorp OnDemand, and UnitedHealthcare-related entities.

A gap at the brand level should not automatically be translated into a parent-company market-share conclusion.

Market Share Must Be Defined Sector by Sector

Questions This Section Answers

• What market-share denominator should be used?
• Can one market-share definition work across every sector?
• Why is denominator matching essential?

There is no universal market-share denominator for this research.

The denominator must match the economic activity represented by the prompt universe as closely as possible.

Examples include:

A recommendation-share comparison is meaningful only if the AI prompt universe and the market-share denominator describe reasonably similar competitive activity.

For example, comparing AI recommendations for "best online savings accounts" with a bank's total corporate assets would be a poor match.

Comparing those recommendations with a relevant deposit or account-share measure could be more defensible.

This matching problem is one reason Article 11 does not publish a universal AI Share Index for the current 25-company panel.

The market-side denominators still need to be standardized sector by sector.

The Proposed Research Variables

The eventual longitudinal study can separate several related variables rather than relying on one ratio.

1. AI Recommendation Share

AI Recommendation Share = Company recommendations / Total qualifying company recommendations in the defined competitive set

This is a competitive AI-side measure.

2. Real-World Market Share

Market Share = Company economic activity / Total economic activity in the matched real-world market

The exact denominator will vary by sector.

3. Recommendation Share Gap

Recommendation Share Gap = AI Recommendation Share - Market Share

This expresses the difference in percentage points.

4. Recommendation Share Ratio

Recommendation Share Ratio = AI Recommendation Share / Market Share

A ratio above 1.0 means AI recommendation share exceeds current market share. A ratio below 1.0 means AI recommendation share trails current market share.

AIVO's revised framework uses a related ratio called the AI Share Index, based on Organic Win Rate divided by market share. The LLM Authority Index framework should treat that as prior art while independently validating its own definitions, prompt universe, platform aggregation, and economic outcomes.

5. Gap Momentum

A static gap may be less informative than its movement.

Gap Momentum = Current Recommendation Share Gap - Prior Recommendation Share Gap

This asks whether the AI-versus-market discrepancy is widening or narrowing over time.

6. Cross-Platform Gap Breadth

A company may have a positive aggregate gap because of one platform while underperforming on the others.

Cross-platform breadth should therefore record how many AI platform families show the same directional discrepancy.

7. Gap Persistence

A gap that appears for one month and disappears the next may be noise.

A gap that persists across several monthly observations may deserve more attention.

The planned cross-platform portability study will help determine how much weight should be placed on multi-platform agreement.

Why the Direction of Change May Matter More Than the Level

The research program should distinguish level from momentum.

A company could have AI recommendation share above market share for years without gaining any economic share. That persistent gap might simply reflect how AI systems represent the category.

A potentially more informative pattern would be:

1. AI recommendation share begins rising;
2. the recommendation-share gap widens;
3. the movement persists across platforms;
4. branded search or direct traffic begins improving;
5. acquisition metrics strengthen;
6. market share or revenue share later rises.

The reverse sequence could also be tested.

That is a stronger leading-indicator hypothesis than assuming a static ratio automatically predicts growth.

The core AI Commercial Momentum Hypothesis is already structured around change over time rather than static visibility alone.

What the Current V0 Panel Can and Cannot Tell Us

The current V0 public-company panel measures recommendation coverage, presence, rank, platform breadth, uncertainty, and cleaning sensitivity across 25 mapped public parents.

It does not yet contain a standardized real-world market-share denominator for every company.

That means the current panel can identify companies with unusual AI recommendation momentum, but it cannot yet responsibly label them as having positive or negative recommendation-share gaps relative to actual economic share.

For example, the current panel shows strong positive recommendation-coverage momentum for Axos Financial and MetLife.

It also shows directional negative recommendation momentum for companies including Bank of America, Coinbase, Goldman Sachs / Marcus, and UWM Holdings.

Those are AI-side changes only.

Without a synchronized and economically matched market-share denominator, we should not infer that Axos or MetLife are gaining real-world share, or that the negative candidates are losing it.

Sector Research Will Be the Natural Testing Ground

The market-share question is easier to study inside sectors than across an arbitrary collection of public companies.

Sector research allows the prompt universe, competitive set, and real-world denominator to be better aligned.

The upcoming sector studies provide that structure:

• 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

Within a sector, we can ask more precise questions.

For banks, does recommendation share for deposit-oriented prompts precede changes in consumer deposit acquisition or related share measures?

For insurers, does recommendation share for a specific line of insurance precede changes in policies, premiums, or quote activity?

For mortgage companies, does AI recommendation share precede origination-volume share?

For brokerage or crypto, does recommendation share precede funded accounts, active users, assets, or transaction volume?

For healthcare, does AI recommendation share correspond to a consumer-facing line that is economically meaningful to the parent?

Those are testable questions.

The Most Important Confounders

Any recommendation-share-versus-market-share study can produce misleading results if the following factors are ignored.

Distribution differences

Real-world market leaders may rely on retail locations, brokers, wholesalers, enterprise relationships, physician referrals, employers, or installed-base advantages that AI prompts do not capture.

Prompt-population differences

AI share depends on what users ask. A prompt set concentrated on affordability, digital convenience, or a niche use case can structurally favor different companies than the overall market.

Geography

A national market-share statistic may be inappropriate if the company's offering or AI recommendation strength is regional.

Product mix

A parent company may dominate one product and be weak in another. The market denominator must correspond to the prompt cluster.

Price and capacity

AI systems can recommend a company more frequently than the company can economically serve. Capacity constraints, underwriting standards, inventory, eligibility rules, pricing, or regulation can block conversion.

Brand architecture

AI may recommend a subsidiary or product name while market-share data is reported at the parent level.

Model and source bias

AI recommendation distributions can be influenced by retrievability, third-party publishers, training data, prompt wording, model updates, and platform-specific behaviors.

These are not reasons to abandon the concept. They are reasons to design the study carefully.

How We Plan to Validate the Gap

Questions This Section Answers

• How will we test whether the recommendation-share gap predicts anything?
• What outcomes should be measured first?
• What would count as failure?

The validation should proceed in stages.

Stage 1: Standardize sector-specific market-share denominators

For each sector, define a denominator that matches the economic behavior implied by the prompt universe.

Document the source, period, units, company mapping, and any exclusions.

Stage 2: Freeze AI recommendation share before future outcomes

Recommendation share should be measured using predeclared prompts and platform families at time T.

Future market-share data must not be used to modify the earlier AI signal.

Stage 3: Compare static level and momentum separately

Test whether the absolute gap matters.

Separately test whether changes in the gap matter.

This avoids assuming that a long-standing structural discrepancy is automatically predictive.

Stage 4: Test intermediate commercial outcomes

Before testing stock returns, compare the gap with outcomes closer to the proposed mechanism:

• branded search;
• web or app engagement;
• direct traffic;
• quote activity;
• account openings;
• customer acquisition;
• transaction volume;
• relevant revenue or segment growth.

Stage 5: Test market-share change

Ask whether positive gaps or widening positive gaps precede later market-share gains, and whether negative gaps precede later losses.

Use sector controls and matched business definitions.

Stage 6: Test incremental value

Compare a baseline model with an AI-augmented model.

The AI gap only becomes useful if it improves forward prediction beyond variables already known at the time, such as prior market-share trend, price, distribution, search demand, web traffic, company size, category growth, and seasonality.

Stage 7: Move to financial outcomes only after commercial validation

Only if the gap predicts economically meaningful business outcomes should the research proceed to analyst revisions, earnings surprises, or sector-relative returns.

This follows the validation order defined in Can AI Search Visibility Predict Revenue Growth?.

What Would Falsify the Recommendation-Share Gap Thesis?

The recommendation-share gap should not be protected from failure.

Evidence against the thesis would include:

• positive gaps do not precede stronger commercial outcomes;
• negative gaps do not precede weaker outcomes;
• gap momentum adds no information beyond prior market-share trend;
• the apparent relationship disappears out of sample;
• the relationship exists only because prompt sets favor certain company types;
• results reverse under reasonable market-share definitions;
• brand-level gaps do not translate to economically meaningful parent-company exposure;
• one platform drives the result while others disagree;
• static AI recommendation bias explains the gap better than commercial momentum;
• or the effect is too small to improve forecasting or due diligence.

These failure conditions are consistent with the broader precommitted framework in What Would Prove the AI Commercial Momentum Hypothesis Wrong?.

Recommendation Share vs. Market Share Is Different From AI Visibility Market Divergence

The two concepts should not be conflated.

AI Visibility Market Divergence compares a future validated AI-derived commercial signal with investor expectations, such as analyst estimates, guidance, or market-implied growth.

This article compares AI recommendation share with real-world competitive market share.

The distinction is:

Recommendation Share Gap
= AI competitive position versus current economic competitive position

AI Visibility Market Divergence
= validated AI-implied commercial trajectory versus financial-market expectations

A company could have a positive recommendation-share gap while investors already expect rapid growth.

Another company could have recommendation share aligned with market share but still show divergence from analyst expectations because its AI momentum is changing quickly.

These are separate research layers.

Could the Gap Eventually Become a Valuation Input?

Possibly, but only after several earlier questions are answered.

AIVO's September 2026 correction is instructive. Its original LLM Equity Valuation framework converted AI recommendation behavior too directly into a valuation-style figure and later withdrew the published Gruns numbers after identifying unit and reach assumptions that overstated the economic interpretation.

The revised framework is more disciplined. It compares Organic Win Rate with market share, separates visibility from conditional conversion, and explicitly states that the central purchase-allocation assumption remains unvalidated.

That sequence supports a conservative rule for this research:

Do not convert AI recommendation-share gaps into dollar value until the relationship between recommendation share and real economic outcomes has been validated.

Even if the gap predicts future commercial share, valuation would still require margins, capital intensity, duration, competitive response, cost of capital, and market expectations.

That is why the LLM Authority Index program separates:

AI measurement
-> commercial validation
-> financial expectations validation
-> market divergence
-> possible future valuation use

The backtesting framework will specify how those later stages should be evaluated.

Investor Interpretation Matrix

What This Research Does Not Mean

The recommendation-share gap does not currently establish:

• that AI systems cause market-share changes;
• that recommendation share equals purchase share;
• that a positive gap means future revenue acceleration;
• that a negative gap means future revenue decline;
• that a company's stock is mispriced;
• that one AI platform represents the whole AI discovery market;
• or that current market share should mechanically converge toward AI recommendation share.

The current research measures AI behavior and creates a framework for testing the economic relationship later.

That distinction is essential.

Methodology Requirements for a Future Recommendation-Share Study

A credible study should disclose at least the following:

1. Prompt universe: the exact unbranded commercial questions used and how they were governed over time.
2. Competitive set: which brands or entities were eligible in each category.
3. Recommendation definition: what qualifies as a recommendation rather than a mention or citation.
4. Platform families: which AI systems were measured and how they were aggregated.
5. Observation period: when the AI data was captured.
6. Market-share denominator: the exact real-world economic measure used for comparison.
7. Market-share period: the date or reporting period corresponding to that denominator.
8. Entity mapping: how products, subsidiaries, legacy brands, and public parents were reconciled.
9. Failure handling: how extraction failures and unavailable responses were treated.
10. Duplicate handling: how overlapping prompt exports or repeated captures were treated.
11. Uncertainty: confidence intervals or other measures of sampling variability.
12. Persistence: whether the gap appeared once or across repeated periods.
13. Outcome horizon: which later commercial outcomes were tested and when.
14. Baseline model: what traditional variables the AI signal had to improve upon.
15. Out-of-sample procedure: how future periods or held-out companies were used to prevent retrospective overfitting.

These requirements build on the current AI Commercial Momentum methodology.

Why This Could Matter to Investors If It Works

If the recommendation-share gap proves predictive, it could become useful because it would measure something different from ordinary financial reporting.

Financial statements report what has already happened.

Market share usually arrives with a delay and often at quarterly or annual frequency.

AI recommendation behavior can be measured repeatedly and at high granularity across prompt clusters.

A validated gap might therefore help investors identify:

• brands gaining disproportionate AI-mediated consideration before share data catches up;
• incumbents whose AI consideration is weakening despite strong existing market position;
• product niches where emerging competitors are being surfaced unusually often;
• sectors where AI systems systematically reshape the competitive consideration set;
• and cases where AI recommendation momentum and conventional market-share momentum disagree.

The important phrase is if it works.

This series is designed to establish that answer prospectively rather than assume it.

What We Will Test Next

The next steps for the recommendation-share research program are:

1. define comparable market-share denominators for each sector;
2. compute competitive AI recommendation share from governed unbranded prompt sets;
3. preserve the static gap and monthly gap momentum separately;
4. evaluate cross-platform breadth and persistence;
5. test the gap first against branded search, traffic, acquisition, and sector-specific commercial measures;
6. test whether the gap predicts future changes in market share or revenue share;
7. compare baseline models with AI-augmented models;
8. preserve out-of-sample results, including failures;
9. proceed to analyst revisions and market outcomes only after commercial validation.

The AI Investor Signal Tracker will preserve the dated AI-side history while those downstream observations accumulate.

Related LLM Authority Index Research

• Can AI Search Signal Future Revenue Growth? The AI Commercial Momentum Hypothesis
• How We Measure AI Commercial Momentum: Methodology for AI Investor Signals
• AI Investor Signal Tracker: Public Company AI Recommendation Momentum
• Can AI Search Visibility Predict Revenue Growth? How We Plan to Test the Relationship
• AI Visibility Market Divergence: Can AI Recommendation Momentum Reveal Information Not Yet Reflected in Investor Expectations?
• Does Cross-Platform AI Visibility Matter? Measuring Recommendation Portability and Platform Concentration Risk
• How Investors Could Backtest AI Search Signals Against Revenue, Analyst Estimates and Stock Performance
• 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

Sources and External Prior Art

1. Alejandro Medina Sandin, AI Meets Antitrust: How Large Language Models Reshape Market Concentration, April 2026. https://ssrn.com/abstract=6627078
2. AIVO Editorial Board, Correcting LLM Equity Valuation: why we withdrew our Gruns figures and what replaces them, September 23, 2026. https://www.aivojournal.org/correcting-llm-equity-valuation-why-we-withdrew-our-gruns-figures-and-what-replaces-them/

Research Note

This article is part of the LLM Authority Index AI Investor Signals research program. It is designed to create a dated, auditable record of the research framework before future commercial and financial outcomes are known.

Nothing in this article is investment advice. The framework does not currently identify undervalued or overvalued securities and does not recommend buying, selling, or holding any security.