AI Visibility Market Divergence: Can AI Recommendation Momentum Reveal Information Not Yet Reflected in Investor Expectations?

Research status: Prospective research framework. AI recommendation momentum has not been validated as a predictor of 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

AI Visibility Market Divergence is the research concept that a validated AI commercial signal could eventually be compared with what analysts, financial models, or market prices appear to expect about a company's future growth.

The idea is straightforward: if AI recommendation momentum eventually proves to contain forward-looking commercial information, then the most useful investor question may not be whether AI visibility is rising or falling by itself. The more useful question may be whether the AI-derived commercial signal is moving differently from prevailing expectations.

For example, a company could show strengthening AI recommendation momentum while analyst revenue expectations remain flat. Another company could show weakening AI recommendation momentum while consensus expectations remain strong. Those situations could become research candidates for deeper investigation if, and only if, the AI signal first demonstrates predictive value in prospective testing.

LLM Authority Index is not currently publishing valuation conclusions from this framework. The July-September 2026 panel measures the AI side of the equation. It does not yet establish the financial side.

The research sequence is deliberately ordered:

AI recommendation measurement
-> AI Commercial Momentum
-> validated relationship with commercial outcomes
-> comparison with analyst and market expectations
-> AI Visibility Market Divergence research
-> possible future valuation use only if the earlier stages survive testing

The underlying theory is defined in Can AI Search Signal Future Revenue Growth? The AI Commercial Momentum Hypothesis. The prospective validation protocol is defined in Can AI Search Visibility Predict Revenue Growth?. The current AI-side signals are preserved in the AI Investor Signal Tracker.

What Is AI Visibility Market Divergence?

Questions This Section Answers

• What does AI Visibility Market Divergence mean?
• Is AI Visibility Market Divergence the same thing as saying a stock is mispriced?
• Why compare AI recommendation momentum with investor expectations?

AI Visibility Market Divergence is a proposed research framework for comparing a validated AI-derived commercial signal with contemporaneous financial expectations.

The framework begins with two different information layers.

The first layer is the AI commercial signal. This includes measurements such as recommendation coverage, recommendation change, platform breadth, persistence, competitive recommendation share, and eventually any validated relationship between those variables and later commercial outcomes.

The second layer is the expectations signal. This includes information such as analyst revenue forecasts, analyst estimate revisions, forward growth expectations, company guidance, valuation multiples, and other variables that represent what the market already appears to expect.

The divergence question is then:

Does the AI-derived commercial signal imply a different directional outlook from the expectations already visible in financial markets?

That is not the same as saying the market is wrong.

A divergence can exist for many reasons. The AI signal may be noisy. Analyst forecasts may incorporate information that AI recommendation data misses. The company may have business segments that are unrelated to the measured AI prompts. The stock price may already reflect information not visible in consensus estimates. Macroeconomic, regulatory, capital-structure, margin, or valuation effects may dominate the commercial signal.

For that reason, AI Visibility Market Divergence should initially be treated as a research discrepancy to investigate, not as proof of mispricing.

The concept becomes credible only if the AI-side measure first passes the validation tests published in What Would Prove the AI Commercial Momentum Hypothesis Wrong?.

Why Divergence Could Matter More Than AI Visibility Alone

An investor usually does not earn informational value simply by knowing that a company is doing well or poorly.

What matters is often the difference between what is happening and what was already expected to happen.

A company can report strong growth and still disappoint investors if expectations were even stronger. A company can report weak growth and still exceed expectations if the market anticipated something worse. The same principle applies to any new alternative-data input.

If AI recommendation momentum eventually proves to contain forward-looking commercial information, then its potential value would likely depend on whether that information is already reflected in conventional expectations.

That creates a more precise research sequence:

1. Measure AI recommendation behavior.
2. Determine whether changes in the AI signal precede measurable commercial outcomes.
3. Estimate the commercial information contained in the AI signal.
4. Compare that information with analyst and market expectations known at the same date.
5. Test whether the divergence predicts later estimate revisions, earnings surprises, or sector-relative returns.

This is why AI Search as Alternative Data treats AI search as a candidate information layer rather than a standalone investment model.

The Four Basic AI and Expectations States

The market-divergence framework can be understood without assigning a stock rating.

These four states are descriptive research categories only.

They do not establish valuation, expected returns, or whether investors should prefer one company over another.

The important question is whether the disagreement itself contains incremental information.

If the AI signal and consensus expectations disagree but later outcomes systematically move toward the AI signal, the divergence could become interesting.

If the AI signal and consensus disagree but later outcomes systematically follow consensus instead, that would count against the value of the AI signal.

If the two measures simply move together, AI recommendation data might still be useful for marketing intelligence but contribute little new information to financial research.

Questions This Section Answers

• What should represent investor expectations in the divergence model?
• Should stock price be the only market-expectations input?
• Why are analyst revenue revisions especially important?

How We Plan to Measure Investor Expectations

There is no single perfect measure of investor expectations.

A stock price compresses many expectations into one number, but it does not reveal which assumptions are driving that price. Analyst estimates are more explicit, but analysts can disagree, revise at different speeds, and focus on different forecast horizons. Company guidance can be informative, but management may frame expectations strategically and not every company provides the same detail.

The market-divergence framework should therefore use multiple expectation measures rather than one universal benchmark.

1. Analyst revenue-growth expectations

Consensus revenue estimates are one of the most direct financial variables to compare with a hypothesized commercial-demand signal.

If AI recommendation momentum is supposed to contain information about future customer consideration and demand, then revenue expectations are a more natural early target than stock returns.

The main questions are:

• Does positive AI recommendation momentum precede upward revenue-estimate revisions?
• Does negative AI recommendation momentum precede downward revisions?
• Does adding the AI signal improve the ability to predict the direction or magnitude of revisions?

2. Analyst estimate revisions

The direction and pace of estimate revisions can serve as a measure of how professional expectations are changing over time.

This is especially relevant to the divergence concept because the signal is not merely trying to explain the current consensus. It is trying to test whether the AI-side measurement contains information before consensus changes.

Academic research has long examined whether forecast revisions are predictable and whether information reaches analysts and markets gradually. A 2019 Journal of Accounting, Auditing and Finance paper by Jung, Keeley, and Ronen found analyst forecast revisions were predictable in their sample and studied whether that predictability supported implementable strategies. A 2026 SSRN working paper titled The Analyst Information Frontier reports that public-signal themes can predict later analyst updates and that some public information remains associated with returns after conditioning on analyst beliefs. These findings are not evidence for AI recommendation data specifically, but they support the broader research question of whether new public signals can precede changes in professional expectations.

Sources:

• https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2991938
• https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6962647

3. Company guidance and management expectations

Where comparable guidance exists, the research can ask whether AI momentum changes before management raises, lowers, or maintains revenue expectations.

Guidance is not interchangeable with analyst consensus, and companies differ in how frequently and precisely they guide. It should therefore be a separate expectation variable rather than blended into one score.

4. Forward valuation measures

Forward valuation multiples can help describe how aggressively the market is pricing expected future earnings or cash flow.

Possible measures include:

• forward price-to-earnings;
• enterprise value to forward revenue;
• enterprise value to EBITDA;
• price-to-sales;
• free-cash-flow yield;
• sector-relative valuation percentiles;
• and company-relative historical valuation ranges.

These variables require careful sector treatment. A bank, health insurer, software company, mortgage lender, and crypto platform should not be compared using the same valuation framework.

5. Market-implied expectations

In more advanced work, it may be possible to reverse-engineer the growth assumptions implied by valuation models.

Instead of asking whether a stock is "cheap" or "expensive," the research question would be:

What revenue, margin, or cash-flow trajectory appears necessary to justify the current market value under a stated model, and how does that compare with the commercial trajectory implied by validated AI data?

That could eventually become one of the cleanest versions of the divergence test because it compares two explicit forward assumptions.

It is also one of the easiest places to create false precision, so it should come later in the research sequence.

Why Analyst Revisions Are a Better Early Target Than Valuation Labels

Analyst revisions provide an intermediate test between commercial outcomes and stock prices.

The proposed mechanism is:

AI recommendation momentum
-> consumer or buyer consideration
-> branded search, traffic, acquisition, or sales
-> company commercial performance
-> analyst revenue revisions
-> earnings expectations and market response

If AI recommendation data contains useful information, one plausible result is that estimate revisions begin moving in the same direction after the AI signal is measured.

This is testable without making a valuation claim.

A 2026 SSRN paper examining firm macroeconomic expectations and analyst forecasts found that analysts issued more favorable forecasts for firms with more optimistic disclosed macroeconomic expectations and that the measured expectations predicted future real activity and profitability in the authors' sample of Chinese A-share companies. The study is not about AI recommendation visibility, and its population is different from the U.S. public-company panel here. It is relevant only as an example of research testing whether a separately measured information signal can precede analyst forecast changes.

Source:

• https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6574840

For LLM Authority Index, the analogous question is whether externally measured AI commercial momentum contains information that later appears in analyst revisions.

That is a stricter standard than showing that AI recommendation coverage correlates with company size, current revenue, or existing analyst optimism.

What the Current V0 Panel Can and Cannot Tell Us

Questions This Section Answers

• Do the current 25-company signals already show market divergence?
• Which companies have the strongest AI-side movements?
• Why are the current labels not valuation signals?

The current public-company panel measures the AI side of the proposed divergence framework.

It does not yet contain a synchronized, prospectively validated market-expectations layer.

Under the V0 rules, the initial 25-company panel contains:

• 2 positive AI divergence candidates;
• 11 negative AI divergence candidates;
• 12 mixed or neutral observations.

The phrase AI divergence candidate in the current tracker refers to unusual AI recommendation movement under the V0 thresholds. It does not mean that the company is already classified as having a divergence from market valuation or analyst expectations.

Selected AI-side examples include:

Those observations are already timestamped in Initial Findings From 25 Public Companies.

The next step is not to label the corresponding stocks.

The next step is to preserve what analysts and financial markets expected at the same point in time, then observe what changed afterward.

For company-by-company updates, the permanent record is maintained in the AI Investor Signal Tracker.

A Proposed AI Visibility Market Divergence Formula

A future quantitative framework could be expressed conceptually as:

AI Visibility Market Divergence = Validated AI-Implied Commercial Trend - Market-Implied Commercial Trend

Both terms require careful construction.

The first term does not exist as a validated financial quantity yet. Today we have measured AI recommendation momentum. We do not yet have a defensible conversion from recommendation change to expected revenue growth.

The second term can be represented in several ways depending on the test:

• analyst consensus revenue growth;
• recent analyst estimate revisions;
• management guidance;
• market-implied growth derived from valuation;
• or a composite expectations model.

Only after the AI signal demonstrates a stable relationship with future commercial outcomes should the first term be calibrated into an expected commercial effect.

Until then, the responsible implementation is a two-column comparison, not a single valuation score:

This structure preserves transparency and prevents a model from hiding fundamentally different inputs inside one composite number.

Questions This Section Answers

• When would a positive divergence become meaningful?
• When would a negative divergence become meaningful?
• What evidence would show that the divergence framework is not useful?

What Would Make AI Visibility Market Divergence Meaningful?

A positive or negative divergence becomes meaningful only after several conditions are met.

Condition 1: The AI signal must first predict something real

Before comparing AI data with valuation or analyst expectations, the AI signal must demonstrate prospective information about a commercial outcome.

That validation process is predeclared in Can AI Search Visibility Predict Revenue Growth?.

If recommendation momentum does not improve forecasts of branded search, traffic, revenue expectations, revenue growth, or another economically meaningful outcome, then building a market-divergence model on top of it would add sophistication without information.

Condition 2: Expectations must be frozen at the same date

A valid divergence test cannot compare an AI signal measured in September with analyst estimates revised months later and pretend both were known at the same time.

The information set must be synchronized.

For a signal date T:

• AI variables must be measured using only data available at T;
• analyst estimates must be the estimates available at T;
• valuation variables must use prices and fundamentals available at T;
• guidance must be limited to what management had already published at T;
• future outcomes must remain unknown.

This prevents look-ahead bias.

Condition 3: Divergence must add information beyond the components

Suppose positive AI momentum predicts later revenue growth, and low consensus expectations also predict positive revisions.

The divergence is useful only if the difference between those information sets contains additional information beyond simply including both variables independently.

That can be tested directly.

A baseline model might include:

• prior revenue growth;
• historical earnings surprise;
• analyst consensus growth;
• recent analyst revisions;
• sector effects;
• valuation;
• branded search or traffic;
• and macro variables.

An AI-enhanced model would add the pre-specified AI variables.

A divergence model would then test whether the interaction or residual difference between AI-implied commercial trend and market expectations improves out-of-sample performance further.

Condition 4: The relationship must survive new periods

A divergence framework should not be accepted because it explains the first 25 companies after the outcome is known.

It must work in later months, new companies, or new sectors that were not used to define the relationship.

That is why the backtesting framework will emphasize walk-forward and out-of-sample evaluation.

Why We Are Not Converting AI Recommendation Share Directly Into Valuation

There is already a useful public example of why this step requires caution.

AIVO published and later corrected an LLM Equity Valuation framework. On September 23, 2026, AIVO said its original formula had overstated the revenue reachable through AI and had compared an annual revenue quantity with a capitalized transaction value. The organization withdrew previously published Gruns valuation figures and revised the methodology around an unbranded Organic Win Rate and an AI Revenue Gap relative to market share.

AIVO's correction is valuable prior art because it demonstrates the danger of moving too quickly from AI recommendation behavior to a dollar valuation.

Source:

• https://www.aivojournal.org/correcting-llm-equity-valuation-why-we-withdrew-our-gruns-figures-and-what-replaces-them/

LLM Authority Index is therefore separating the research stages.

We first measure AI recommendations. Then we test whether the measures predict later commercial outcomes. Then we test whether they precede changes in financial expectations. Only after those relationships survive prospective validation would it make sense to explore whether the signal belongs in a valuation framework.

AI Recommendation Share vs. Market Share Is a Related but Different Divergence

A second type of divergence compares AI recommendation share with actual market share.

That question is explored separately in AI Recommendation Share vs. Market Share.

The distinction matters:

• AI recommendation share vs. market share asks whether AI systems recommend a company more or less often than its real-world category position might suggest.
• AI Visibility Market Divergence asks whether a validated AI-derived commercial trend differs from what financial markets or analysts appear to expect about the future.

A company could over-index in AI recommendations relative to market share but still have financial expectations that already assume rapid growth.

Another company could have average AI recommendation share but a sharp positive change in recommendation momentum while analyst expectations remain unchanged.

Those are different research situations and should not be compressed into one score.

Cross-Platform Breadth and Persistence Should Affect Divergence Confidence

A one-platform spike should carry less weight than a persistent multi-platform change.

The V0 methodology already tracks six platform families separately:

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

The later divergence model should preserve that information rather than averaging it away.

A stronger research candidate would generally have:

• measurable recommendation movement;
• directional uncertainty bounds;
• breadth across several AI platforms;
• persistence across multiple monthly observations;
• reasonable robustness to prompt subsets and cleaning choices;
• economically meaningful exposure to the public parent;
• and a synchronized divergence from expectations.

The role of cross-platform behavior is examined in Does Cross-Platform AI Visibility Matter?.

Persistence matters because a one-month movement could reflect model updates, retrieval changes, prompt noise, or temporary information shocks. A divergence that survives several months would be more interesting than an isolated observation, although persistence itself still would not prove financial importance.

Parent-Company Exposure Is Essential

AI systems often recommend products, brands, subsidiaries, or divisions rather than the public parent itself.

That creates a major risk in market-divergence research.

Examples in the current V0 panel include:

• UFB Direct within Axos Financial;
• Marcus within Goldman Sachs;
• Coinbase Wallet within Coinbase Global;
• CVS Pharmacy within CVS Health;
• Labcorp OnDemand within Labcorp;
• UnitedHealthcare consumer brands within UnitedHealth Group.

If a tracked brand represents only a small part of consolidated economics, a large AI movement may have little effect on parent-company revenue.

The divergence framework should therefore incorporate an economic exposure layer wherever practical.

Potential variables include:

• segment revenue share;
• segment operating income share;
• customer-acquisition contribution;
• strategic growth importance;
• geographic contribution;
• and whether the measured product is economically material to the public parent.

Without that layer, a product-level AI signal can be incorrectly interpreted as a whole-company financial signal.

Public Markets, Private Equity, and Venture Capital Use Different Expectation Benchmarks

The market-divergence idea can extend beyond public equities, but the expectations benchmark changes by asset class.

Public equities

Possible expectations inputs include:

• analyst revenue and EPS estimates;
• estimate revision breadth;
• forward valuation multiples;
• company guidance;
• option-implied information where appropriate;
• and market-implied growth assumptions.

Private equity

A private-equity investor may compare AI commercial momentum with:

• management forecasts;
• diligence assumptions;
• customer-growth models;
• transaction multiples;
• lender cases;
• and the acquisition case presented by the seller.

Venture capital and growth equity

A venture investor may compare AI recommendation momentum with:

• ARR growth expectations;
• category-growth assumptions;
• customer acquisition trends;
• fundraising valuation;
• forward revenue multiples;
• and management projections.

The basic concept is the same in all three cases:

Does independently measured AI commercial momentum point in a materially different direction from the expectations used to value the company?

The answer is not automatically actionable, but the discrepancy may become a diligence question if the AI signal is validated.

What Would Disprove the Market-Divergence Framework?

The market-divergence concept should be rejected or narrowed if future testing shows any of the following persistent patterns:

1. AI commercial momentum does not predict later commercial outcomes. Without a validated first-stage signal, there is nothing meaningful to compare with market expectations.
2. AI momentum moves only after analyst estimates change. In that case it may be a lagging reflection rather than an early signal.
3. Divergence adds no value beyond the separate AI and expectations variables. The composite concept would be unnecessary.
4. Apparent divergence is caused by one platform, one prompt set, or one cleaning choice. The signal would be too fragile for financial interpretation.
5. Product-level signals fail after adjusting for parent-company economic exposure. The stock-level mapping would be invalid.
6. Out-of-sample results disappear. Historical fit without forward replication would not support the framework.
7. The direction changes materially when reasonable expectation measures are substituted. A robust divergence concept should not depend on one arbitrary consensus or valuation definition.

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

How We Would Backtest AI Visibility Market Divergence

The later backtest should preserve the real information sequence.

For each historical signal date T:

1. Freeze the AI recommendation metrics available at T.
2. Freeze analyst estimates, guidance, valuation, and conventional digital data available at T.
3. Use only models and rules that would have been available at T.
4. Calculate the AI-versus-expectations divergence using a versioned formula.
5. Observe future estimate revisions, revenue results, earnings surprises, and later sector-relative returns.
6. Move forward to the next period without refitting the historical observation using future information.

Evaluation should focus on whether the divergence improves:

• analyst revision-direction accuracy;
• revenue-growth prediction;
• revenue-surprise classification;
• EPS-surprise classification;
• cross-sectional ranking;
• information coefficient;
• and, only later, sector-neutral return analysis.

Detailed procedures will be documented in How Investors Could Backtest AI Search Signals Against Revenue, Analyst Estimates and Stock Performance.

Why the Historical Publication Record Matters

The value of this research program depends partly on preserving what was known when each signal was observed.

If the project later finds that AI recommendation momentum predicted analyst revisions or commercial outcomes, the result will be more credible if the underlying AI signals, methodology, thresholds, and failure conditions were published before the outcomes were known.

If the signal fails, the historical record should make that failure visible too.

That is why the project now includes:

• the original AI Commercial Momentum Hypothesis;
• the initial 25-company findings;
• the methodology;
• the falsification framework;
• the permanent signal tracker;
• the definition of an AI Investor Signal;
• the measurement-separation framework;
• and the prospective revenue-validation protocol.

The market-divergence framework is the next layer, not a replacement for those earlier stages.

What This Does Not Mean

This article does not establish that:

• AI recommendation momentum predicts analyst revisions;
• AI recommendation momentum predicts revenue growth;
• AI recommendation momentum predicts earnings surprises;
• AI recommendation momentum predicts stock returns;
• analyst estimates are inefficient;
• current prices fail to reflect AI-related information;
• a positive AI signal means a stock is attractive;
• or a negative AI signal means a stock is unattractive.

It establishes only a research framework for asking whether a validated AI commercial signal eventually contains information that differs from prevailing financial expectations.

The current V0 labels are measurements of AI recommendation behavior, not conclusions about market pricing.

Methodology Summary

The AI side of the framework currently uses the V0 methodology documented in How We Measure AI Commercial Momentum.

Key features include:

• matched normalized prompts on the same platform family across periods;
• explicit not-mentioned company rows in the denominator;
• exclusion of explicit extraction failures;
• collapse of response-identical cross-vertical duplicate exports in the primary panel;
• manual public-parent rollups for mapped entities;
• cell-weighted recommendation-coverage changes;
• prompt-clustered exploratory uncertainty intervals;
• six-platform breadth measurement;
• no-dedupe and capture-average sensitivity checks;
• and predeclared magnitude, interval, and platform-breadth thresholds for the current candidate labels.

The expectations side has not yet been standardized into one production model. That is intentional. The research will first test multiple expectation measures separately before deciding whether a composite expectations index is justified.

Limitations

The market-divergence framework currently has several important limitations.

The AI signal is not yet financially validated

This is the largest limitation. A divergence between an unvalidated AI signal and analyst expectations is simply a difference between two numbers until the AI side demonstrates commercial information value.

The current time series is short

The core investor prototype begins with roughly three months of AI observations. That is insufficient to establish persistence, cyclicality, or a stable financial relationship.

Analyst expectations are imperfect proxies for market expectations

Consensus estimates are visible and testable, but investors can hold expectations that differ from published analyst forecasts.

Market prices contain more information than consensus estimates

A stock can move before analyst estimates change. That is one reason future work should test both estimate revisions and market outcomes rather than treating consensus as the market itself.

Valuation models require assumptions

Reverse-engineered market-implied growth depends on discount rates, margins, terminal assumptions, capital intensity, and other variables. Any future valuation-based divergence must show those assumptions explicitly.

Parent-company exposure can be incomplete

Some current AI observations represent brands, products, or divisions. Their relevance to consolidated financial outcomes varies by company.

AI platforms are not identical

Recommendation behavior can differ materially across ChatGPT, Gemini, Google AI Mode, Google AI Overviews, Microsoft Copilot, and Perplexity. Platform breadth and persistence must remain visible.

Expectations can adjust quickly

Even if AI recommendation data initially contains information, the market may incorporate that information before a practical investor can use it. Predictive value and usable timing are separate questions.

What We Will Test Next

The market-divergence research will proceed only after the commercial validation layer begins producing enough forward observations.

The planned sequence is:

1. Continue monthly AI recommendation measurements using fixed and governed prompt universes.
2. Freeze analyst revenue and EPS expectations at each AI signal date.
3. Record estimate-revision direction over subsequent 30-day and 90-day windows.
4. Compare AI momentum with branded search and web/app engagement where reliable data is available.
5. Compare AI momentum with subsequent reported revenue growth and revenue surprise.
6. Estimate whether AI variables improve baseline models out of sample.
7. Only after that step, calculate formal AI-versus-expectations divergence variables.
8. Test whether divergence predicts later analyst revisions or financial surprises better than either input alone.
9. Evaluate sector-specific models where the relationship appears economically plausible.
10. Examine sector-relative stock returns only after the commercial and expectations tests are sufficiently mature.

If these tests fail, the market-divergence concept should be narrowed or abandoned rather than relabeled.

Related LLM Authority Index Research

• Can AI Search Signal Future Revenue Growth? The AI Commercial Momentum Hypothesis
• Can AI Recommendation Momentum Identify Emerging Company Trends? Initial Findings From 25 Public Companies
• AI Search as Alternative Data: Could AI Recommendations Become a Leading Indicator for Investors?
• How We Measure AI Commercial Momentum: Methodology for AI Investor Signals
• What Would Prove the AI Commercial Momentum Hypothesis Wrong?
• AI Investor Signal Tracker: Public Company AI Recommendation Momentum
• What Is an AI Investor Signal? How AI Search Visibility Could Become Financial Alternative Data
• AI Recommendations vs. Mentions vs. Citations: Which AI Visibility Metrics Could Matter Most to Investors?
• Can AI Search Visibility Predict Revenue Growth? How We Plan to Test the Relationship
• AI Recommendation Share vs. Market Share: Could the Gap Reveal Emerging Commercial Strength or Weakness?
• How Investors Could Backtest AI Search Signals Against Revenue, Analyst Estimates and Stock Performance

Research References

• Jung, Michael J., Jessica Keeley, and Joshua Ronen. The Predictability of Analyst Forecast Revisions. Journal of Accounting, Auditing and Finance. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2991938
• Taheri Hosseinkhani, Nima. The Analyst Information Frontier. SSRN, 2026. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6962647
• Guo, Peng, Fuwei Jiang, Mengru Li, and Can Yang. What Do Firms Signal to Analysts? Macroeconomic Expectations and Analyst Forecasts. SSRN, 2026. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6574840
• AIVO Journal. 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 Status Summary

The current evidence supports measuring AI recommendation momentum and prospectively testing whether it predicts future commercial outcomes.

It does not yet support treating a difference between AI recommendation momentum and current financial expectations as proof that the market has missed information.

That is the purpose of AI Visibility Market Divergence research: not to declare that the market is wrong, but to test whether a validated AI commercial signal eventually identifies information that becomes visible in analyst expectations, reported results, or market outcomes later.