AI Search as Alternative Data: Could AI Recommendations Become a Leading Indicator for Investors?

Research status: Exploratory longitudinal research. AI recommendation momentum 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

Publication date: October 3, 2026

Answer Capsule

AI search data could become a form of alternative data for investors if changes in AI recommendations reliably precede changes in consumer consideration, digital demand, revenue expectations, or other economically meaningful outcomes. That relationship has not yet been established.

The investment research opportunity is not simply to measure whether a company appears in ChatGPT, Gemini, Google AI Mode, Google AI Overviews, Microsoft Copilot, or Perplexity. The more useful question is whether persistent changes in unbranded AI recommendation behavior contain incremental information about a company's future commercial trajectory.

LLM Authority Index is testing that question prospectively through the AI Commercial Momentum Hypothesis. The initial public-company panel shows measurable gains and losses in recommendation coverage across 25 mapped public parents, but those observations are not yet predictive evidence. The complete starting panel is published in Initial Findings From 25 Public Companies.

The central test is simple: does a frozen AI recommendation signal measured today improve our ability to explain or predict what happens later, after controlling for information investors already have?

If the answer is yes, AI recommendation momentum may become a useful alternative-data input. If the answer is no, the signal should remain an AI-search measurement rather than an investment signal.

Key Findings and Research Position

The current evidence supports studying AI search as alternative data, but it does not support treating AI visibility as a standalone investment model.

• In the initial LLM Authority Index public-company panel, 2 of 25 public parents qualified as positive AI divergence candidates, 11 qualified as negative AI divergence candidates, and 12 were mixed or neutral under the current V0 rules.
• The two positive candidates were Axos Financial, with a +19.7 percentage-point change in matched recommendation coverage, and MetLife, with a +11.3-point change.
• Several negative candidates showed declines across five or six platform families, including Bank of America, Goldman Sachs / Marcus, Coinbase, Chime, American Express, and Ally.
• The median public-parent difference between the primary cleaned signal and the no-dedupe sensitivity calculation was 0.0 percentage points, suggesting that the broadest current movements are not solely artifacts of one duplicate-handling choice.
• External research now shows that AI systems are participating in real purchase journeys. A 2026 Marketing Science study analyzed 973 e-commerce websites, more than 50,000 ChatGPT-referred transactions, and more than 164 million transactions from traditional channels. The study found that organic LLM referral traffic remained small but produced measurable commercial outcomes. Read the peer-reviewed study.
• NielsenIQ reported in September 2026 that 51% of U.S. consumers had used at least one AI-powered tool to support shopping in the prior month, with AI-powered product recommendations the most widely used application in its tracker. Read the NIQ release.
• G2's 2026 B2B software buyer research reported that 51% of surveyed buyers start software research with an AI chatbot more often than Google, and 71% use AI chatbots somewhere in the software research process. Read G2's Answer Economy report.
• NIQ and Similarweb announced a measurement collaboration intended to connect AI-driven discovery with consumer intent, traffic, and sales conversion, which illustrates that the commercial measurement layer around AI discovery is beginning to develop. Read the announcement.
• At the same time, AIVO publicly corrected its LLM Equity Valuation methodology in September 2026 after determining that an earlier formula overstated AI-reachable revenue. That correction is an important warning against converting recommendation share directly into valuation before the economic relationship is validated. Read AIVO's correction.

The evidence therefore supports a research program, not a conclusion.

Investor Research Action Matrix

Questions This Section Answers

• What is alternative data in investing?
• Could AI search visibility qualify as alternative data?
• What would make AI recommendation data useful to investors?

What Would Make AI Search Data "Alternative Data" for Investors?

In investment research, alternative data generally refers to information outside traditional company filings, financial statements, management guidance, and conventional market data that may help explain or anticipate business performance.

Examples can include web traffic, app usage, credit-card spending, search demand, job postings, pricing, satellite imagery, shipping activity, customer reviews, and other behavioral or operational signals.

AI search data fits naturally into that conceptual category because it measures a new form of digital market behavior: which companies AI systems surface, compare, and recommend when users ask commercially meaningful questions.

But being nontraditional does not automatically make a dataset useful.

For AI search data to become credible investment alternative data, it must demonstrate at least four properties:

1. Economic relevance. The measured behavior must plausibly connect to real consumer or business decisions.
2. Repeatability. The signal must survive repeated measurement rather than changing randomly with model output.
3. Time ordering. The AI signal must be measured before the financial or commercial outcome it is supposed to help explain.
4. Incremental information. The signal should add information beyond variables investors already observe, such as prior growth, search demand, traffic, valuation, seasonality, and sector trends.

The AI Commercial Momentum Hypothesis is designed around those requirements.

The narrower What Is an AI Investor Signal? article will define the terminology used throughout this research program, while How We Measure AI Commercial Momentum documents the actual measurement rules.

Why AI Recommendations Could Matter Economically

Questions This Section Answers

• How can an AI recommendation affect a company's commercial funnel?
• Is AI search already connected to real buying behavior?
• Why might recommendation momentum matter more than simple visibility?

An AI recommendation is potentially important because modern answer engines do more than retrieve information. They can reduce a market from dozens of possible companies to a short list of candidates.

That creates a potential commercial chain:

Unbranded commercial question
→ AI-generated comparison or recommendation
→ consideration set
→ branded search, site visit, app visit, marketplace visit, or direct navigation
→ purchase or customer acquisition
→ revenue and earnings

This chain is plausible, and pieces of it are now observable, but the entire chain has not been proven as a stable investment relationship.

The strongest current evidence is at the behavioral and transaction level.

The 2026 Marketing Science study of ChatGPT referrals found measurable sessions, transactions, revenue, conversion rates, and engagement across 973 e-commerce websites. The researchers examined more than 50,000 ChatGPT-referred transactions during a 12-month dataset and found that organic LLM traffic had higher conversion rates and revenue per session than paid social, but lower performance than most other traditional digital channels. The authors also emphasized an important limitation: last-click attribution can understate channels that contribute primarily to discovery. See the study.

That limitation is especially relevant to investor research. If a user asks an AI system which companies to consider, then later navigates directly to a brand, searches for it on Google, opens an app, or purchases through a retailer, the AI system may have influenced the decision without receiving the final referral attribution.

NIQ's September 2026 Agentic Commerce Tracker adds a consumer-behavior perspective. NIQ reported that 51% of U.S. consumers had used at least one AI-powered tool to support shopping in the prior month, with adoption spanning discovery, comparison, evaluation, and final purchase decisions. See NIQ's findings.

For B2B software, G2's 2026 survey of more than 1,000 buyers reported that 51% begin research in an AI chatbot more often than in Google, 71% rely on AI chatbots somewhere in the process, and 69% said they chose a different vendor than originally planned based on AI chatbot guidance. See G2's report.

Those findings do not prove that AI recommendation share predicts public-company revenue. They do establish why recommendation behavior is commercially relevant enough to test.

Recommendation Data Is Not the Same as Citation Data

A major risk in AI-search measurement is treating every observable output as the same thing.

They are not.

A company can be:

• cited without being mentioned prominently;
• mentioned without being recommended;
• recommended without ranking first;
• ranked highly on one platform and absent on another;
• described positively but not selected;
• frequently cited while losing commercial recommendation share.

For investment research, these differences matter because the hypothesized economic mechanism is most directly connected to consideration and recommendation, not simply to whether a source URL appeared in an answer.

LLM Authority Index therefore keeps at least six measurement layers separate:

1. citations;
2. mentions or share of voice;
3. recommendations;
4. rank or ordering;
5. sentiment;
6. downstream business and financial outcomes.

The dedicated AI Recommendations vs. Mentions vs. Citations study will examine which of these measures appears most economically informative over time.

Why Unbranded Commercial Prompts Matter

If the objective is to measure commercial momentum, asking an AI model a branded question can contaminate the signal.

A prompt such as "Is MetLife a good insurance company?" already places MetLife in the consideration set.

A prompt such as "What are the best life insurance companies for a 45-year-old parent?" asks the model to construct the consideration set itself.

That difference matters.

Unbranded prompts let us observe whether a company is being selected organically within a relevant commercial category. They are therefore closer to a behavioral market-share concept than branded awareness queries.

This does not mean every unbranded prompt is equally valuable. Prompt intent, category fit, consumer type, geography, product availability, seasonality, and prompt wording can all change the result.

The solution is not to find one supposedly perfect prompt. It is to define a stable prompt universe, measure it repeatedly, and compare like with like.

That is why the primary V0 signal uses matched prompt and platform cells rather than comparing unrelated prompt pools across months. The full design is documented in How We Measure AI Commercial Momentum.

What the Initial 25-Company Panel Shows

The current public-company panel is useful because it demonstrates that AI recommendation behavior can move materially at the company level even over a relatively short observation window.

The complete starting panel is preserved in Initial Findings From 25 Public Companies. The purpose of publishing it now is not to claim predictive power. It is to freeze the signal before later outcomes are known.

Selected examples include:

These measurements should be interpreted exactly as written: they show changes in matched AI recommendation coverage. They do not establish changes in revenue, customer acquisition, earnings quality, intrinsic value, or future stock returns.

That distinction is central to this research program.

Questions This Section Answers

• Can AI search predict revenue growth today?
• Can investors trade directly on AI recommendation momentum today?
• What would need to happen before the signal could be used more confidently?

Why the Signal Is Not Yet Predictive Evidence

No. LLM Authority Index has not established that AI search predicts future revenue growth.

No. The current signal should not be treated as a standalone trading rule.

The initial dataset gives us a timestamped explanatory variable. It does not yet give us a validated dependent relationship.

To move from interesting measurement to investment-grade alternative data, we need prospective evidence that a signal observed at time T is related to an economically meaningful outcome observed at time T+1, and that the relationship survives controls, alternative specifications, and out-of-sample tests.

The planned validation sequence is described in Can AI Search Visibility Predict Revenue Growth? and How Investors Could Backtest AI Search Signals Against Revenue, Analyst Estimates and Stock Performance.

At minimum, we would want to know whether current AI recommendation momentum predicts any of the following better than a reasonable baseline model:

• future branded search demand;
• future direct or organic traffic;
• future app or marketplace engagement where observable;
• analyst revenue-estimate revisions;
• reported revenue growth;
• sales surprise;
• earnings surprise;
• sector-relative stock returns.

The signal becomes more interesting only if it contributes incremental explanatory or predictive value after controlling for ordinary information such as prior growth, valuation, sector, seasonality, search demand, traffic, and other known variables.

The Most Important Distinction: Commercial Signal vs. Market Signal

A company can gain commercial momentum without being an attractive stock.

A company can lose commercial momentum while still being attractively valued.

That is why the long-term architecture of this research separates three layers:

Layer 1: AI Commercial Momentum

What is changing in AI-mediated commercial consideration?

Layer 2: Financial Outcomes

Does that change later appear in revenue, customer acquisition, earnings, or analyst expectations?

Layer 3: Market Expectations

Was the commercial change already reflected in the stock's valuation and consensus expectations?

Only after those layers are separated can we meaningfully study AI Visibility Market Divergence.

A positive AI signal paired with a low valuation might eventually prove different from the same AI signal paired with an extremely high valuation. Likewise, a negative commercial signal might matter differently when consensus expectations are already pessimistic.

Those are valuation questions, not AI-visibility questions.

The current research is intentionally starting with the visibility and recommendation layer first.

AI Recommendation Share vs. Market Share

Another possible use of AI-search data is to compare a company's share of AI recommendations with its actual market share.

Conceptually, a company that receives substantially more AI recommendation share than its current real-world category share may be over-indexing in AI-mediated consideration. A company receiving much less recommendation share than its existing market position may be under-indexing.

That difference could be commercially interesting.

But it should not be converted automatically into expected revenue or valuation.

AIVO's September 2026 correction illustrates why. AIVO revised its LLM Equity Valuation framework after finding that its earlier formula overstated the revenue reachable through AI and compared financial quantities that should not have been treated as directly equivalent. The revised framework compares unprompted AI recommendation share with market share, but AIVO explicitly notes that the assumption linking recommendation share to AI-influenced purchases remains unvalidated. Read the correction.

LLM Authority Index will examine this question separately in AI Recommendation Share vs. Market Share.

The sequence matters:

First measure recommendation share.
Then observe future economic outcomes.
Then test whether the relationship is stable.
Only then consider valuation applications.

Cross-Platform Breadth May Be More Informative Than a Single-Platform Spike

A recommendation signal can look powerful on one platform and disappear on another.

That is why the V0 system tracks ChatGPT, Gemini, Google AI Mode, Google AI Overviews, Microsoft Copilot, and Perplexity separately before summarizing company-level momentum.

This relates directly to the LLM Authority Index Persistence-Portability Gap.

A signal can be:

• persistent over time but concentrated on one AI platform;
• broadly portable across platforms but short-lived;
• both persistent and portable;
• neither persistent nor portable.

For investment research, the third condition is the most intuitively interesting, but that intuition itself needs validation.

For example, the initial MetLife signal improved on five of six platform families, while Bank of America declined on all six. Those are broader cross-platform movements than a one-engine change. However, even a six-platform movement still does not prove anything about future revenue until the financial outcome is observed.

The dedicated cross-platform AI visibility study will examine portability and concentration directly.

Persistence May Matter More Than One-Month Movement

Alternative data becomes more useful when it distinguishes durable change from noise.

One month of recommendation movement could reflect:

• model updates;
• retrieval changes;
• temporary news coverage;
• prompt sensitivity;
• content freshness;
• data extraction variation;
• genuine shifts in the information ecosystem;
• genuine shifts in commercial relevance.

Those mechanisms are not equivalent.

The research program therefore needs repeated observations using a fixed or carefully versioned prompt universe. If a company gains recommendation coverage for several consecutive months across several platforms, the signal may deserve more weight than a one-month spike.

This is another reason the AI Investor Signal Tracker is central to the project. The tracker is intended to preserve the historical sequence rather than continually replacing older observations with the newest snapshot.

Where AI Search Alternative Data May Be Most Useful

The economic relevance of AI recommendation behavior is unlikely to be uniform across sectors.

It should be strongest where several conditions hold:

• consumers or business buyers actively research providers before purchasing;
• recommendations materially narrow the choice set;
• digital discovery is important to customer acquisition;
• switching between brands is realistic;
• purchase decisions are not determined almost entirely by geography, regulation, long-term contracts, or commodity pricing;
• the measured product or brand represents a meaningful share of the listed parent's economics.

That suggests potentially stronger applications in areas such as:

• banking and consumer financial products;
• insurance;
• fintech;
• lending and mortgages;
• software and technology purchasing;
• direct-to-consumer products;
• travel and hospitality;
• selected healthcare services;
• other high-consideration consumer categories.

The initial sector work will examine bank stocks, insurance stocks, fintech, brokerage and crypto, mortgage and lending stocks, and healthcare stocks.

By contrast, AI recommendation behavior may have much weaker economic relevance for businesses whose revenue is driven mainly by commodity prices, regulated returns, long-cycle industrial contracts, or other forces that have little connection to consumer or buyer discovery.

That heterogeneity should be measured rather than assumed away.

Existing Market Activity Shows That Investor-Facing AI Visibility Is Already Emerging

The idea that AI outputs matter to investors is not unique to this project.

In June 2026, 5W AI Communications released an IPO AI Visibility Index covering 25 recent and pending U.S. IPO candidates across ChatGPT, Claude, Gemini, Perplexity, and Google AI Overviews. Its framework focused on recognition, accuracy, source control, and answer quality. That is primarily an investor-relations and narrative-readiness use case, not a revenue-prediction model. See the index announcement.

Presenc AI has also published an investor-relations use case focused on how AI assistants describe public companies, including business models, segments, risks, competitive positioning, and financial accuracy. Again, that is primarily narrative and information-quality monitoring rather than commercial-momentum forecasting. See Presenc AI's IR use case.

These examples matter because they show that AI visibility is already becoming relevant to investor-facing workflows.

The LLM Authority Index research question is different:

Can unbranded commercial recommendation behavior become predictive alternative data about the business itself, rather than merely describing how investors see the company?

That is a substantially harder claim, which is why longitudinal validation is required.

NIQ and Similarweb Point Toward the Measurement Layer We Need

In September 2026, NielsenIQ and Similarweb announced work on a new solution intended to connect AI-driven discovery with consumer intent, product content, AI-driven traffic, and sales conversion. Read the NIQ and Similarweb announcement.

That measurement direction is highly relevant to the AI Investor Signals thesis.

The central challenge is attribution.

If AI systems influence product discovery but the eventual conversion occurs through direct navigation, Google, an app, a marketplace, a store, or an offline channel, referral traffic alone will miss some portion of the commercial effect.

A robust investor signal therefore may need to combine:

• AI recommendation behavior;
• AI referral traffic;
• branded search demand;
• direct traffic;
• app activity;
• marketplace behavior;
• transaction data;
• company-reported revenue;
• analyst estimate revisions.

The AI signal should be one layer in that stack, not the entire stack.

Questions This Section Answers

• How should investors use AI search data today?
• What would an institutional-quality AI signal require?
• How should AI signals be combined with traditional financial research?

How Investors Could Use AI Search Data Responsibly Today

At its current stage, AI search data is best used as a research-screening and hypothesis-generation signal.

An analyst could use it to identify questions worth investigating:

• Which companies are gaining recommendation share across several platforms?
• Which established brands are losing AI-mediated consideration?
• Are challenger brands appearing more often in unbranded commercial prompts?
• Are recommendation changes persistent for several months?
• Is the movement concentrated in one product line or broad across categories?
• Does the AI signal move before branded search or traffic?
• Does it move before analyst estimates change?
• Does it add anything after conventional financial variables are included?

The correct workflow is therefore:

AI signal
→ investigative question
→ traditional financial and commercial validation
→ prospective outcome tracking

Not:

AI signal
→ automatic stock conclusion

That distinction is what separates an alternative-data research program from a marketing score dressed up as investment analysis.

An institutional-quality version would also require strong provenance, reproducible prompt sets, timestamped observations, extraction-status controls, entity mapping, survivorship controls, versioning, and out-of-sample testing.

What Would Prove the Signal Is Useful?

A useful result does not require AI recommendation momentum to predict stock returns directly.

It might prove valuable at an earlier stage of the information chain.

For example, the research could find that AI recommendation momentum predicts:

• branded search changes but not revenue;
• website traffic changes but not analyst revisions;
• analyst revisions in some sectors but not others;
• revenue surprises only when the signal persists for three months;
• commercial changes only when movement appears across four or more platforms;
• consumer-facing segment outcomes but not consolidated parent-company results.

Any of those would be informative.

The research could also fail completely.

If recommendation changes do not precede meaningful commercial or financial changes, if the relationship disappears after controlling for existing variables, or if the apparent effect cannot be reproduced out of sample, AI recommendation momentum should not be promoted as a financial leading indicator.

That falsification standard is developed in What Would Prove the AI Commercial Momentum Hypothesis Wrong?.

AIVO's Correction Shows Why Time Ordering Matters

AIVO's September 2026 correction is one of the most useful pieces of prior art for this field because it demonstrates the danger of moving too quickly from AI recommendation behavior to valuation.

The revised AIVO framework compares a brand's share of unprompted AI recommendations with its market share. That comparison may be commercially interesting. But AIVO also acknowledged that its earlier valuation formula overstated AI-reachable revenue and withdrew the earlier Grüns figures.

The lesson is broader than one formula.

If a researcher observes a strong AI recommendation share today and a successful company outcome today, it is tempting to convert the association into a value claim.

A better test is prospective:

1. freeze the AI signal today;
2. publish the methodology today;
3. specify the outcomes before observing them;
4. wait for those outcomes;
5. test whether the earlier signal added information;
6. preserve the misses as well as the hits.

That is the structure of the LLM Authority Index AI Investor Signals program.

Methodology Summary

The detailed methodology is maintained separately in How We Measure AI Commercial Momentum. The key V0 principles are summarized here.

Source corpus

The preserved LLM Authority Index research archive contains:

• 68,203 raw observations;
• 310,114 raw citation-array entries;
• 624,698 company-array entries, including explicit not-mentioned company records.

Those are archive preservation counts, not the screened investor sample.

Primary signal

The investor prototype uses company-array fields rather than citation counts as its principal signal source. The primary metric is recommendation coverage, defined as the proportion of eligible matched prompt/platform cells in which a company or mapped public parent received a valid recommendation.

Matched-panel construction

The primary change compares the same normalized prompt on the same platform family in the base month and September 2026. July is preferred as the base month. Travelers uses August because July was unavailable in its matched panel.

Failure handling

The preserved archive contains 1,278 observations explicitly flagged as extraction failures. Those observations are excluded and are not treated as legitimate zero-visibility responses.

Duplicate handling and sensitivity

The primary panel collapses repeated exports inferred to represent the same response state. A no-dedupe sensitivity calculation is also maintained. A separate repeated-capture sensitivity analysis averages distinct retained response variants within prompt/platform/month cells.

Public-parent mapping

Brands and products are mapped to listed parents where appropriate. A parent is counted as recommended when any mapped tracked entity is recommended in a matched cell.

This creates an important limitation. A row based on Marcus by Goldman Sachs, Coinbase Wallet, CVS Pharmacy, Labcorp OnDemand, or another product-level entity should not automatically be interpreted as a whole-company demand measure.

Cross-platform breadth

The system separately tracks ChatGPT, Gemini, Google AI Mode, Google AI Overviews, Microsoft Copilot, and Perplexity. Platform breadth is used as a descriptive measure of portability, not as proof of financial relevance.

Confidence intervals

The current intervals are exploratory. Matched-cell changes are aggregated within normalized prompts, and a normal-approximation interval is calculated across prompt-cluster means. They are not causal confidence intervals.

Limitations

This research has substantial limitations that should remain visible as the series develops.

1. Three months is too short for investment validation

July through September 2026 is enough to create a baseline, not enough to establish predictive performance through multiple economic and earnings cycles.

2. The company panel is not a representative equity index

The initial 25 public parents were derived from companies available in the existing commercial AI-search corpus. They should not be treated as a random sample of U.S. public companies.

3. AI usage differs by sector

AI recommendation behavior may be economically meaningful in high-consideration consumer categories while having much less relevance in businesses driven by other factors.

4. Recommendation output can change for reasons unrelated to company fundamentals

Model updates, retrieval systems, source freshness, news cycles, prompt construction, and platform-specific policies can change recommendations.

5. Brand-to-parent mapping can overstate economic relevance

A consumer-facing brand may represent only a portion of a diversified public parent.

6. Recommendation coverage is not market share

It measures behavior inside a defined prompt and platform universe. It is not real-world sales share unless separately demonstrated.

7. AI referral traffic may understate AI influence

Consumers can discover a brand through AI and convert through another channel, making attribution difficult.

8. Public market prices incorporate many variables

Even a genuine commercial leading indicator may not predict stock returns if the information is already priced in or overwhelmed by other factors.

What We Will Test Next

The next stage is to connect frozen AI signals to future outcomes without changing the original observations after the fact.

Planned validation layers include:

Near-term commercial indicators

• branded search changes;
• website traffic changes;
• app or marketplace engagement where available;
• other observable demand proxies.

Financial expectations

• 30-day and 90-day analyst revenue-estimate revisions;
• EPS estimate revisions;
• changes in consensus expectations.

Reported outcomes

• subsequent revenue growth;
• sales surprise;
• EPS surprise;
• segment-level results where the measured brand maps to a specific business line.

Market outcomes

• sector-relative returns;
• excess returns over longer windows;
• whether any relationship survives valuation and risk controls.

The long-term test is not whether we can find a few examples that look convincing. It is whether the signal works systematically, prospectively, and out of sample.

The formal testing framework will be documented in How Investors Could Backtest AI Search Signals Against Revenue, Analyst Estimates and Stock Performance.

Why Publishing the Signal Before the Outcome Matters

Alternative-data research is especially vulnerable to retrospective storytelling.

If we waited until a company reported unexpectedly strong growth and then searched backward for an AI metric that had risen beforehand, the evidence would be much weaker.

Publishing now creates a record of:

• the hypothesis;
• the metric definition;
• the company universe;
• the original signal direction;
• the confidence classification;
• the known limitations;
• the validation criteria.

Future updates can then show whether each observation was useful, irrelevant, misleading, or sector-specific.

The AI Investor Signal Tracker will preserve that history rather than replacing prior observations with the newest interpretation.

That record may ultimately be more valuable than any one month's ranking.

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
• 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 Visibility Market Divergence: Can AI Recommendation Momentum Reveal Information Not Yet Reflected in Investor Expectations?
• AI Recommendation Share vs. Market Share: Could the Gap Reveal Emerging Commercial Strength or Weakness?
• 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
• The Persistence-Portability Gap: Why AI Citation Authority Persists Over Time but Fragments Across Platforms

External References

1. Kaiser, Maximilian, and Christian Schulze. "ChatGPT Referrals to E-Commerce Websites: How Do LLMs Compare Against Traditional Channels?" Marketing Science, published online April 21, 2026. https://pubsonline.informs.org/doi/10.1287/mksc.2025.0489
2. NielsenIQ. "Majority of U.S. Consumers Now Use AI to Shop, NIQ Finds." September 24, 2026. https://nielseniq.com/global/en/news-center/2026/majority-of-u-s-consumers-now-use-ai-to-shop-niq-finds/
3. NielsenIQ and Similarweb. "NIQ and Similarweb Advance Agentic Commerce Measurement for the AI Shopping Era." September 2, 2026. https://nielseniq.com/global/en/news-center/2026/niq-and-similarweb-advance-agentic-commerce-measurement-for-the-ai-shopping-era/
4. G2. "The Answer Economy: How AI Search Is Rewiring B2B Software Buying." 2026. https://learn.g2.com/g2-2026-ai-search-insight-report?price=FREE
5. 5W AI Communications. "IPO AI Visibility Index." June 2026. https://www.morningstar.com/news/pr-newswire/20260601ny72492/5w-ai-communications-releases-the-ipo-ai-visibility-index-the-companies-going-public-are-losing-the-first-question-buyers-ask
6. Presenc AI. "AI Visibility Monitoring for Investor Relations Teams." Updated April 23, 2026. https://presenc.ai/use-cases/ai-visibility-for-investor-relations
7. AIVO. "Correcting LLM Equity Valuation: Why We Withdrew Our Grüns 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 Disclosure

This article reports exploratory research into AI recommendation behavior. It is not investment advice, a securities recommendation, a valuation opinion, or a forecast of any company's revenue, earnings, or stock performance. Company names and ticker symbols are used to identify entities included in the research panel. The AI Investor Signals project is designed to test whether the observed measurements have any future economic relationship, including the possibility that no reliable relationship exists.