AI Investor Signal Tracker: Public Company AI Recommendation Momentum

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

Current methodology version: V0

Initial observation window: July through September 2026

Current 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

Tracker role: Permanent index for the LLM Authority Index AI Investor Signals research program

Answer Capsule

The AI Investor Signal Tracker is the permanent public index for LLM Authority Index research on whether changes in AI recommendation visibility can become a useful commercial or financial signal.

The current V0 public-company panel contains 25 mapped public parents. Using the predeclared V0 rules, the initial July-September 2026 snapshot contains:

• 2 positive AI divergence candidates;
• 11 negative AI divergence candidates;
• 12 mixed or neutral observations;
• 7 High confidence AI-measurement classifications;
• 16 Medium confidence classifications;
• 2 Exploratory classifications.

The strongest positive recommendation-coverage changes in the initial public panel were Axos Financial at +19.7 percentage points and MetLife at +11.3 percentage points. The largest negative changes included UWM Holdings at -11.5 points, Goldman Sachs / Marcus at -11.3 points, Labcorp OnDemand at -10.0 points, Chime at -9.4 points, CVS Health at -9.4 points, Coinbase at -9.1 points, and Bank of America at -8.6 points.

Those are measurements of AI recommendation behavior, not stock ratings. They do not establish that a company is undervalued, overvalued, likely to beat earnings, likely to miss earnings, or likely to outperform or underperform its peers.

The purpose of this tracker is to preserve the signal history, link each company to its sector and company research page, document methodology changes, and make later validation auditable. The underlying research question is defined in The AI Commercial Momentum Hypothesis, the initial complete panel is published in Initial Findings From 25 Public Companies, and the measurement rules are documented in How We Measure AI Commercial Momentum.

Current AI Investor Signal Snapshot

Questions This Section Answers

• Which public companies currently have the strongest positive or negative AI recommendation momentum?
• How many companies are positive, negative, or mixed in the initial tracker?
• Does a positive or negative AI signal mean the stock should be bought or sold?

The initial V0 tracker contains two positive AI divergence candidates, eleven negative AI divergence candidates, and twelve mixed or neutral observations. These classifications are based on AI recommendation movement only.

A positive AI divergence candidate must have at least a +5 percentage-point recommendation-coverage change, an exploratory 95% interval with a lower bound above zero, and at least four AI platform families improving. A negative AI divergence candidate requires at least a -5 percentage-point change, an interval with an upper bound below zero, and at least four platform families worsening.

No tracker classification is a buy, sell, hold, valuation, or stock-performance conclusion. The signal is being frozen now so that later financial and commercial outcomes can be compared against what was observed before those outcomes were known.

Initial 25-Company Public Tracker

How to Read the Tracker

Questions This Section Answers

• What does recommendation coverage measure?
• What does the confidence label mean?
• Why can a company move substantially without becoming a positive or negative candidate?

Recommendation coverage is the proportion of eligible matched prompt-platform cells in which the company or mapped public parent received a valid recommendation. It is not the same as citation share, mention share, sentiment, market share, revenue share, or stock performance.

The primary movement is the change between the base month and September 2026 across the same normalized prompts on the same platform families. July is the preferred base month in V0. Travelers uses August because July is unavailable in its matched panel.

The confidence label describes confidence in the AI measurement, not confidence in a future financial outcome.

• High requires at least 200 matched cells, all six platform families, a directional exploratory 95% interval, and both cleaning-sensitivity differences within 2 percentage points.
• Medium requires at least 100 matched cells and at least five platform families.
• Exploratory applies to rows that do not meet the High or Medium rules.

A company can have a large positive or negative point estimate and still remain mixed or neutral. The V0 candidate rule also requires directional uncertainty and broad cross-platform support. For example, Labcorp OnDemand declined by 10.0 percentage points, but the interval crosses zero and the row is Exploratory. It is therefore not labeled a negative AI divergence candidate.

The complete technical definition is in How We Measure AI Commercial Momentum.

Sector Research

The tracker is intentionally organized around both companies and sectors. AI recommendation behavior may prove more useful in some commercial categories than others, so sector context is necessary before interpreting a company-level movement.

Banking

See Bank Stocks and AI Search: Which Banks Are Gaining or Losing AI Recommendation Visibility?.

The banking and consumer-finance subset includes companies such as Axos Financial, Bank of America, Citigroup, Ally Financial, Goldman Sachs / Marcus, American Express, and related financial-services entities in the broader panel.

Insurance

See Insurance Stocks and AI Search: Which Insurers Are Gaining or Losing AI Recommendation Visibility?.

The insurance-related panel includes MetLife, Principal Financial Group, Allstate, Travelers, Lincoln Financial, Corebridge Financial, Trupanion, and other insurance-linked companies.

Fintech, Brokerage and Crypto

See Fintech, Brokerage and Crypto Stocks: What AI Recommendation Momentum Shows So Far.

This group includes Webull, Coinbase, Chime, Upstart, Happen / LendingClub, and adjacent digital-finance brands.

Mortgage and Lending

See Mortgage and Lending Stocks in AI Search: Initial Recommendation Momentum Signals.

This group includes PennyMac, UWM Holdings, Upstart, Happen / LendingClub, and related lending companies.

Healthcare

See Healthcare Stocks and AI Search: Which Companies Are Gaining or Losing AI Recommendation Visibility?.

This group includes Molina Healthcare, Cigna, UnitedHealth Group through UnitedHealthcare entities, CVS Health through CVS Pharmacy, and Labcorp through Labcorp OnDemand.

Why the Tracker Exists

The tracker serves a different purpose from a one-time article.

The dated research articles freeze what was known at publication. The tracker provides the current navigation layer and longitudinal status of the program.

That distinction is important.

If a company that had a positive AI signal later experiences weaker commercial results, the original article should not be rewritten to make the initial signal disappear. If a negative signal later reverses, the dated record should still show that the original decline existed.

The tracker can add the new month, update the current state, and link back to the historical observation.

This creates a public record of:

1. what the AI recommendation signal looked like at the time;
2. which methodology version produced it;
3. how the signal changed in subsequent months;
4. what commercial and financial outcomes followed;
5. whether the hypothesis became stronger, weaker, narrower, or unsupported.

The full initial snapshot is permanently documented in Initial Findings From 25 Public Companies.

Research Program Navigation

Questions This Section Answers

• Where should an investor start with this research?
• Which articles explain the signal versus test the signal?
• Where will future validation results be published?

For readers new to the series, the recommended research sequence is:

1. Start with Can AI Search Signal Future Revenue Growth? The AI Commercial Momentum Hypothesis.
2. Review the frozen first panel in Initial Findings From 25 Public Companies.
3. Read AI Search as Alternative Data for the broader investor-data framework.
4. Review How We Measure AI Commercial Momentum before interpreting individual company movements.
5. Read What Would Prove the AI Commercial Momentum Hypothesis Wrong? for the precommitted failure conditions.
6. Use this tracker to navigate company and sector pages.
7. Follow the planned revenue-growth validation study, cross-platform study, and backtesting framework as the longitudinal record develops.

The conceptual distinction between different AI measurements will also be addressed in AI Recommendations vs. Mentions vs. Citations, while the eventual question of whether AI behavior differs from market expectations will be treated separately in AI Visibility Market Divergence.

What Will Change on This Page Over Time?

This tracker is intended to evolve. The historical research articles are not.

As new months are added, this page can be updated with:

• current recommendation coverage;
• latest month-over-month movement;
• trailing three-month movement;
• cross-platform breadth;
• persistence of prior signals;
• changes in confidence classification;
• new company or sector pages;
• links to monthly research updates;
• later downstream validation fields when they become available.

The current V0 classification should remain visible as the initial baseline even after later methodology versions are introduced.

Methodology version history

A methodology revision can improve future measurement, but it should not erase the original historical record.

Monthly Update Archive

The first tracker state covers July through September 2026 and is being published in October 2026.

Future monthly research updates will use a consistent title and slug convention so the history remains easy to retrieve.

Planned format:

• AI Investor Signals: October 2026 Public Company Recommendation Momentum Update
• AI Investor Signals: November 2026 Public Company Recommendation Momentum Update
• AI Investor Signals: December 2026 Public Company Recommendation Momentum Update

The corresponding canonical URL convention will be:

https://llmauthorityindex.com/resources/ai-investor-signals/[month]-[year]-ai-investor-signals-public-company-recommendation-momentum-update

Once a monthly update is published, this tracker should link to it without changing the historical content of prior monthly articles.

Persistence Matters More Than a One-Month Spike

One of the central questions in this research is whether an AI recommendation movement persists.

A large one-month gain could reflect:

• a genuine shift in how multiple AI systems evaluate a company;
• a model update;
• a changing citation or retrieval environment;
• prompt-specific volatility;
• a temporary news cycle;
• or measurement noise.

The same is true of a one-month decline.

That is why the research program will distinguish movement from persistence.

If a signal remains directionally consistent across multiple months and multiple AI platforms, it becomes more interesting as a potential commercial indicator. If it quickly reverses, its investment usefulness may be limited even if the original movement was measured correctly.

The cross-platform side of this problem connects to the broader LLM Authority Index Persistence-Portability Gap, which separates whether authority persists over time from whether it transfers across AI platforms.

Brand-to-Parent Mapping Remains a Major Interpretation Constraint

Some tracker rows represent a direct public-company brand. Others represent a consumer-facing product, operating company, subsidiary, or legacy brand that has been rolled to a public parent.

Examples include:

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

These mappings are necessary for public-equity analysis, but they create an economic-exposure question.

A strong product-level AI signal may matter greatly to a parent whose economics depend heavily on that product. The same signal may matter much less to a highly diversified parent where the measured consumer brand is only a small part of consolidated revenue.

The Happen row requires an additional transition caveat because the underlying measured entity is the legacy LendingClub brand during a 2026 rebrand period. A decline in legacy-brand recommendation visibility could represent commercial deterioration, brand migration, AI staleness, or some combination of those factors.

The tracker therefore should be read as a map of measured AI recommendation behavior, not as a one-to-one map of parent-company revenue exposure.

What the Tracker Does Not Measure Yet

The current tracker does not establish:

• revenue growth;
• earnings growth;
• revenue surprise;
• EPS surprise;
• analyst estimate revisions;
• branded search growth;
• website or app traffic growth;
• customer acquisition changes;
• market share changes;
• intrinsic value;
• expected stock return;
• excess return relative to a sector benchmark.

Those are downstream validation targets.

The purpose of the tracker is to freeze the AI-side measurement first. That sequencing prevents future financial outcomes from influencing which historical AI observations are highlighted.

What We Will Test Next

The research program will test progressively harder downstream outcomes.

Stage 1: Search and digital behavior

We will evaluate whether AI recommendation momentum precedes changes in observable consumer behavior such as branded search, site traffic, app engagement, or other category-appropriate digital activity where reliable data is available.

Stage 2: Analyst expectations and reported fundamentals

We will test whether AI recommendation movement contains information about later analyst revenue revisions, reported revenue growth, revenue surprise, EPS surprise, or other company-specific financial outcomes.

The study design is described in Can AI Search Visibility Predict Revenue Growth?.

Stage 3: Market relevance

Only after the commercial relationship is tested will it make sense to ask whether the AI signal adds information about market expectations or future sector-relative returns.

That research is separated into AI Visibility Market Divergence, AI Recommendation Share vs. Market Share, and How Investors Could Backtest AI Search Signals.

What Would Make This Tracker Useful?

The tracker becomes more useful if the historical AI measurements eventually show one or more repeatable relationships with future outcomes.

Potentially useful findings could include:

• persistent AI recommendation gains preceding increases in branded search;
• cross-platform losses preceding weaker web or app demand;
• recommendation momentum improving before analyst revenue estimates rise;
• sector-specific relationships where consumer AI discovery matters more to purchasing behavior;
• stronger predictive value from recommendation measures than from citation or mention measures;
• stronger signal quality from multi-platform persistence than from one-platform spikes.

But the tracker can also become useful by showing where the hypothesis fails.

If AI recommendation momentum has little relationship with later commercial outcomes, that is an important result. If it works in direct-to-consumer finance but not diversified healthcare, that is a narrower result. If it predicts search behavior but not revenue, that is a different result again.

The precommitted failure framework is published in What Would Prove the AI Commercial Momentum Hypothesis Wrong?.

Limitations

The current tracker has several important limitations.

1. Short history. The core longitudinal archive covers only about three months. That is not enough to validate a forecasting model.
2. Prompt-population differences. Prompt universes are not identical across all verticals and months. Matched prompt-platform cells reduce this problem but do not eliminate it.
3. Extraction quality. Explicit extraction failures are excluded, but missing upstream requests and silent losses cannot always be reconstructed.
4. Entity resolution. The 25 public-parent mappings are manually curated and should eventually be governed through a formal entity master.
5. Business exposure. Product and subsidiary signals may not represent the economics of the entire parent company.
6. Platform instability. AI systems change models, retrieval policies, ranking behavior, and answer formats over time.
7. Outcome validation is not complete. The tracker measures AI behavior before proving that the behavior predicts financial outcomes.
8. Sector concentration. The initial panel is concentrated in consumer-facing financial, insurance, healthcare, mortgage, brokerage, and related categories.
9. Current intervals are exploratory. The V0 95% interval is a transparent prompt-clustered approximation, not a causal confidence interval.
10. Sentiment is not in the V0 classification. Sentiment fields are retained but not used in the watch-category rules because investment-use calibration has not been validated.

These limitations are discussed in more detail in the AI Commercial Momentum methodology.

Related LLM Authority Index Research

• Can AI Search Signal Future Revenue Growth? The AI Commercial Momentum Hypothesis
• Initial Findings From 25 Public Companies
• AI Search as Alternative Data
• How We Measure AI Commercial Momentum
• What Would Prove the AI Commercial Momentum Hypothesis Wrong?
• What Is an AI Investor Signal?
• AI Recommendations vs. Mentions vs. Citations
• Can AI Search Visibility Predict Revenue Growth?
• AI Visibility Market Divergence
• AI Recommendation Share vs. Market Share
• Does Cross-Platform AI Visibility Matter?
• How Investors Could Backtest AI Search Signals

Research Disclosure

This tracker reports exploratory measurements of AI recommendation behavior. It is research, not personalized investment advice, and the classifications are not recommendations to buy, sell, or hold any security.

The tracker is intentionally being published before the longitudinal financial validation is complete. Future results may support, narrow, modify, or reject the current hypothesis.