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

Research status: Exploratory longitudinal research. AI recommendation momentum has not been validated as a predictor of insurance premium growth, policy sales, membership growth, medical enrollment, analyst revisions, valuation, or stock returns.

Initial observation window: July through September 2026

Insurance and health-plan research slice: 10 mapped public parents

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

Current methodology version: V0

Answer Capsule

The initial LLM Authority Index insurance and health-plan panel shows a mixed pattern in AI recommendation momentum from July to September 2026 rather than one uniform sector direction.

Among ten mapped public-company parents in this research slice, MetLife was the only company that met the V0 positive AI divergence candidate rules. Its recommendation coverage increased 11.3 percentage points, from 27.5% to 38.7%, with an exploratory 95% interval of approximately +4.4 to +18.1 points and positive movement on five of six AI platform families.

Principal Financial Group increased 5.4 points and Molina Healthcare increased 3.7 points, but both remained mixed or inconclusive under the V0 thresholds. Cigna was exactly flat on aggregate recommendation coverage at 35.2% in both periods.

Six companies declined on aggregate recommendation coverage: UnitedHealth Group / UnitedHealthcare -4.6 points, Trupanion -5.4, Travelers -6.0, Allstate -6.4, Lincoln Financial -6.5, and Corebridge Financial -7.6. None of these six rows met every requirement for the V0 negative AI divergence candidate label. In several cases, the exploratory interval crossed zero. In others, platform direction was fragmented rather than consistently negative.

Across the ten-company slice, the median recommendation change was approximately -5.0 percentage points, while the simple mean was approximately -1.6 points. The difference between those summary statistics is important because MetLife's +11.3-point increase pulls the average upward.

These measurements describe AI recommendation behavior only. They do not establish changes in insurance demand, policy count, premiums, claims, underwriting quality, Medicare enrollment, medical membership, earnings, or stock performance.

The research question is narrower and prospective: do persistent, cross-platform changes in unbranded AI recommendation visibility later correspond with measurable changes in customer consideration, branded demand, quote activity, enrollment, policy sales, analyst expectations, or reported financial outcomes?

The theory behind that question is defined in Can AI Search Signal Future Revenue Growth? The AI Commercial Momentum Hypothesis. The company-level measurement rules are documented in How We Measure AI Commercial Momentum, and the complete initial public-company panel is preserved in Initial Findings From 25 Public Companies.

Insurance AI Recommendation Momentum at a Glance

This sector view contains companies with different business models. It includes life insurance, property and casualty insurance, pet insurance, diversified financial services, managed care, pharmacy-benefit exposure, and health-plan brands. It is therefore a research-defined insurance and health-plan slice, not a formal insurance equity index.

The overlap matters. Molina, Cigna, and UnitedHealth Group also appear naturally in the planned Healthcare Stocks and AI Search analysis because the measured consumer journeys sit at the intersection of healthcare and insurance.

Questions This Section Answers

• Which insurer-related companies gained AI recommendation visibility?
• Which companies declined the most?
• Does the negative sector median mean the insurance industry weakened?

Which Insurance Companies Gained or Lost AI Recommendation Visibility?

MetLife: the strongest positive signal in the sector slice

MetLife increased from 27.5% recommendation coverage to 38.7%, a gain of 11.3 percentage points.

The exploratory 95% interval remained above zero, from approximately +4.4 to +18.1 points. Five of six platform families improved:

• ChatGPT: +23.08 pp
• Gemini: +15.38 pp
• Google AI Mode: +8.51 pp
• Google AI Overviews: +9.84 pp
• Microsoft Copilot: +26.32 pp
• Perplexity: -5.41 pp

The row receives a High V0 AI-measurement confidence classification. It has 222 matched cells, six measured platform families, a directional interval, and no observed difference in either the no-dedupe or capture-average sensitivity calculations.

MetLife also illustrates why recommendation coverage and recommendation rank should be separated. Its average observed recommendation rank moved from approximately 3.88 to 4.22, which is numerically worse among the responses where it was recommended, even while recommendation coverage increased materially.

The AI-side conclusion is therefore specific: MetLife was recommended in a larger share of eligible matched commercial prompt-platform cells, but its average rank among observed recommendations did not improve.

That does not establish higher policy sales, premiums, retention, or future earnings.

Principal Financial Group: positive direction, insufficient evidence for a candidate label

Principal Financial Group increased 5.4 percentage points, from 55.9% to 61.3%.

The magnitude exceeds the 5-point watch threshold, but the row does not satisfy the complete positive-candidate rule. Its exploratory interval crosses zero, it contains only 93 matched cells, and only three platforms moved positively.

Platform changes were:

• ChatGPT: +18.18 pp
• Gemini: +8.33 pp
• Google AI Mode: 0.00 pp
• Google AI Overviews: -6.25 pp
• Microsoft Copilot: +41.67 pp
• Perplexity: -14.29 pp

The proper label is therefore Mixed / neutral, with an Exploratory confidence classification.

This distinction is important. A large point estimate is not enough by itself. The V0 framework also requires uncertainty and platform breadth before assigning a directional candidate label.

Molina Healthcare: modest aggregate increase with split platform behavior

Molina Healthcare increased 3.7 points, from 23.5% to 27.2% recommendation coverage.

Its presence coverage increased more, by 6.5 points, from 40.6% to 47.0%.

The platform picture was mixed:

• ChatGPT: -12.50 pp
• Gemini: -7.14 pp
• Google AI Mode: +10.34 pp
• Google AI Overviews: +5.26 pp
• Microsoft Copilot: -4.35 pp
• Perplexity: 0.00 pp

Two platforms improved, three worsened, and one was flat. The exploratory interval crossed zero.

The result is therefore not evidence of broad positive AI momentum. It is a small positive aggregate move with material platform disagreement.

Cigna: aggregate recommendation coverage was unchanged

Cigna is the cleanest example in the sector of an aggregate neutral result.

Recommendation coverage was 35.2% in both the base period and September, producing a 0.0 percentage-point change.

Presence coverage increased modestly by 1.5 points, from 64.3% to 65.9%, while average observed rank improved from approximately 4.38 to 4.03.

Platform movement was balanced:

• ChatGPT: +9.38 pp
• Gemini: 0.00 pp
• Google AI Mode: +0.83 pp
• Google AI Overviews: 0.00 pp
• Microsoft Copilot: -4.08 pp
• Perplexity: -4.26 pp

The current parent mapping combines Cigna and Express Scripts exposure. That mapping helps avoid treating related public-company entities as separate stocks, but it also means the row should not be interpreted as a pure measure of every business operated by The Cigna Group.

UnitedHealth Group: high absolute visibility, modest negative momentum

UnitedHealth Group declined 4.6 percentage points, from 64.5% to 59.8% recommendation coverage.

The row remained one of the highest absolute recommendation-coverage observations in this sector even after the decline.

Its exploratory interval narrowly crossed zero, with an upper bound of approximately +0.3 points, so the V0 result remains inconclusive rather than directional.

The platform changes were:

• ChatGPT: -11.11 pp
• Gemini: -21.88 pp
• Google AI Mode: +4.04 pp
• Google AI Overviews: -4.05 pp
• Microsoft Copilot: 0.00 pp
• Perplexity: -17.65 pp

The row maps UnitedHealthcare, Golden Rule, and UnitedHealthcare Vision consumer-facing entities to UnitedHealth Group. It does not represent Optum or every economic segment of the public parent.

That brand-to-parent limitation is material. A recommendation trend in consumer health-plan prompts may be economically relevant, but it cannot automatically be generalized to consolidated UnitedHealth Group revenue.

Trupanion: aggregate decline, perfectly split platform direction

Trupanion declined 5.4 percentage points, from 53.6% to 48.2% recommendation coverage.

Its platform distribution demonstrates why aggregate momentum and cross-platform portability must remain separate:

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

Three platforms improved and three worsened.

The company therefore had negative aggregate movement but no directional platform majority. This is exactly the type of case discussed in Does Cross-Platform AI Visibility Matter?.

Travelers: August-to-September comparison, not July-to-September

Travelers declined 6.0 percentage points, from 27.2% to 21.2% recommendation coverage.

Travelers is the only public-parent row in the current panel whose matched base month is August 2026, because July is unavailable in the matched panel.

This difference must be preserved in interpretation. Its change is not directly the same time interval as the July-to-September rows.

Platform changes were:

• ChatGPT: +6.25 pp
• Gemini: -26.32 pp
• Google AI Mode: +9.09 pp
• Google AI Overviews: -11.32 pp
• Microsoft Copilot: -28.57 pp
• Perplexity: -8.33 pp

Four of six platforms worsened, but the exploratory interval crossed zero.

Travelers also illustrates the distinction between presence and recommendation. Its presence coverage increased 2.6 points while recommendation coverage declined 6.0 points. An AI visibility metric based only on presence could therefore tell a different story from the recommendation metric.

Allstate: broad negative platform direction, but interval still crosses zero

Allstate declined 6.4 percentage points, from 21.8% to 15.5%.

Five platforms worsened and one was flat:

• ChatGPT: -20.00 pp
• Gemini: 0.00 pp
• Google AI Mode: -4.88 pp
• Google AI Overviews: -2.13 pp
• Microsoft Copilot: -33.33 pp
• Perplexity: -33.33 pp

Despite broad negative platform direction, the exploratory interval extends slightly above zero, to approximately +0.5 points. Under the predeclared V0 rules, the row therefore remains Mixed / neutral.

The public-parent mapping combines Allstate with National General and Direct Auto Insurance exposure. That improves parent-level coverage but also means the result reflects a defined set of consumer-facing brands rather than every business line.

Lincoln Financial: aggregate decline that masks mostly positive platform directions

Lincoln Financial declined 6.5 percentage points, from 30.9% to 24.4%.

But four of six platform families moved positively:

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

This is one of the strongest examples in the entire initial panel of aggregate movement being driven by platform weighting rather than broad directional agreement.

The correct description is therefore not simply that Lincoln Financial lost AI visibility. The more precise conclusion is that its cell-weighted aggregate recommendation coverage declined while platform-level direction was highly fragmented.

Corebridge Financial: largest aggregate decline in the sector slice

Corebridge Financial had the largest aggregate recommendation decline among the ten companies, at -7.6 percentage points.

Recommendation coverage fell from 18.5% to 10.9%, and presence coverage fell 6.7 points.

Platform changes were:

• ChatGPT: -12.50 pp
• Gemini: 0.00 pp
• Google AI Mode: -6.25 pp
• Google AI Overviews: -9.52 pp
• Microsoft Copilot: 0.00 pp
• Perplexity: -10.00 pp

Four platforms worsened and two were flat. The exploratory interval still crossed zero, so the row remains Mixed / neutral under V0.

Why the Sector Median and Mean Tell Different Stories

The ten-company median recommendation change is approximately -5.0 percentage points. The simple mean is approximately -1.6 points.

That gap is not a mathematical curiosity. It reflects the shape of the sector data.

MetLife's +11.3-point gain is large enough to pull the average upward. Principal and Molina also show positive aggregate movement. At the same time, six companies are negative and Cigna is flat.

For a small research panel, a single mean can obscure the distribution.

The more informative description is:

• 3 positive aggregate movers
• 1 exactly flat row
• 6 negative aggregate movers
• 1 positive AI divergence candidate
• 0 negative AI divergence candidates
• 9 mixed or neutral observations

This is meaningfully different from the bank-sector pattern, where several negative rows satisfied the full V0 candidate rule.

Questions This Section Answers

• Why are there no negative AI divergence candidates in this insurance slice?
• Can a company decline more than 5 points and still remain mixed?
• Why does cross-platform direction matter?

Why Most Insurance Declines Remain Mixed or Neutral

The V0 negative AI divergence candidate label requires three conditions simultaneously:

1. recommendation coverage change of -5 percentage points or worse;
2. the exploratory 95% interval must remain below zero;
3. at least four platform families must worsen.

Several insurance-sector rows satisfy the magnitude rule but fail one of the other tests.

Allstate, Travelers, and Corebridge show negative aggregate changes of more than 5 points and at least four worsening platforms, but their intervals still cross zero.

Trupanion exceeds the 5-point decline threshold, but platform direction splits three to three and its interval crosses zero.

Lincoln Financial declines more than 6 points, but four of six platforms actually improve and its interval crosses zero.

That is why none of these rows receives the negative-candidate classification.

This design is intentionally conservative. It prevents a large but noisy point estimate from being treated as equivalent to a large, directional, cross-platform move.

The methodology is documented in How We Measure AI Commercial Momentum.

Cross-Platform Insurance Behavior Is Structurally Important

Insurance is also one of the strongest examples in the broader LLM Authority Index research of why platform-specific AI behavior matters.

The separate citation-source research for high-stakes insurance decisions found a median monthly Top 100 source persistence of 80.2%, but median cross-platform Top 100 portability of only 29.0%, producing an approximate 51.2 percentage-point Persistence-Portability Gap for that citation-source study.

Those figures measure citation-source ecosystems, not insurer recommendation rates. They should not be substituted for the company recommendation metrics in this article.

But the structural lesson is relevant: the insurance information environment can be durable over time while remaining highly fragmented across AI platforms.

That pattern is consistent with the company-level recommendation examples here:

• MetLife rises on five platforms but falls on Perplexity.
• Trupanion splits three positive and three negative.
• Lincoln Financial declines in aggregate even though four platforms move positively.
• Allstate is negative on five platforms and flat on one.

The correct research approach is therefore to retain both the aggregate company signal and the individual platform pattern.

The broader framework is described in the Persistence-Portability Gap and in Does Cross-Platform AI Visibility Matter?.

Recommendation Coverage Is Not the Same as Presence, Rank, or Citation Visibility

Insurance also produces several useful metric-divergence examples.

Travelers gained presence while losing recommendation coverage. MetLife gained recommendation coverage while average observed recommendation rank moved lower. Cigna held recommendation coverage flat while presence rose slightly and average observed rank improved.

Those outcomes reinforce the measurement rule developed in AI Recommendations vs. Mentions vs. Citations:

citation ≠ mention ≠ recommendation ≠ rank ≠ sentiment ≠ purchase ≠ revenue.

A recommendation metric asks whether the company enters the AI-generated consideration set.

A presence metric asks whether it appears in the answer at all.

Rank asks where it appears when recommended.

Citation analysis asks which sources support the answer.

These variables may eventually be modeled together, but they should not be blended simply because they all relate to AI visibility.

Why Insurance May Be an Important Validation Sector

Insurance has several characteristics that make it potentially useful for testing the AI Commercial Momentum Hypothesis.

Consumers frequently research insurers using comparison-style questions that fit naturally into conversational AI:

• best life insurance companies for a particular profile;
• best Medicare or health-plan options;
• affordable auto insurers;
• pet insurance comparisons;
• insurance for older consumers;
• policy options based on price, age, coverage, or specific needs.

These are close to the unbranded high-intent prompt structure used in the current research.

At the same time, the commercial path differs substantially by product.

A life-insurance recommendation may lead to a quote, agent contact, application, underwriting process, and delayed policy issuance. A health-plan recommendation may interact with employer coverage, enrollment windows, Medicare eligibility, network availability, and regulation. Property and casualty decisions can be strongly price-sensitive and state-specific. Pet insurance has different purchase cycles and consumer economics.

That heterogeneity creates both opportunity and risk for validation.

If a relationship exists, it may be stronger in some insurance categories than others. A broad sector average could hide product-specific signal quality.

Future tests should therefore examine not only public-parent outcomes but also the business segment actually represented by the AI prompt universe.

Brand-to-Parent Mapping Is a Major Insurance Limitation

Several rows in this sector are not one-to-one mappings between an AI-visible consumer brand and an entire public company.

Examples include:

• MetLife includes MetLife and MetLife Vision exposure.
• Cigna includes Cigna and Express Scripts.
• UnitedHealth Group is represented through UnitedHealthcare, Golden Rule, and UnitedHealthcare Vision consumer entities.
• Allstate includes Allstate, National General, and Direct Auto Insurance.

The parent rollup is necessary because investors analyze listed companies, not disconnected brand labels.

But parent mapping also creates a financial exposure question:

How much of the public parent's revenue, earnings, customer acquisition, or future growth is actually connected to the consumer-facing entity captured by the AI prompts?

That question must be answered before using the AI signal in financial models.

A strong UnitedHealthcare recommendation signal, for example, cannot automatically be treated as a signal for every UnitedHealth Group segment. Likewise, a consumer insurance brand may represent only part of a diversified financial company.

The prospective validation framework therefore calls for segment-aware economic exposure when possible rather than assuming every brand contributes equally to parent-company value.

AI Recommendation Share vs. Real Insurance Market Share

A second future test is whether a company's competitive AI recommendation share differs from its real-world insurance market share.

That concept is developed in AI Recommendation Share vs. Market Share.

Insurance is particularly sensitive to denominator choice.

A valid real-world market-share denominator might be:

• direct written premium for a defined insurance line;
• covered lives or membership for a defined health-plan market;
• policy count;
• new-business premium;
• Medicare Advantage enrollment;
• pet-insurance policy count;
• or another product-specific commercial measure.

Those denominators are not interchangeable.

For that reason, this article does not publish a universal AI recommendation-share-to-market-share ratio across the ten companies. The sectors and products differ too much for one denominator to be meaningful.

What This Does Not Mean

The initial insurance-sector findings do not show that:

• MetLife will grow revenue faster than the other companies;
• Corebridge, Lincoln, Allstate, Travelers, Trupanion, or UnitedHealth Group will experience weaker financial results;
• companies with negative AI recommendation momentum will lose real-world market share;
• a company with broad platform agreement is a better or worse investment;
• a positive AI divergence candidate is undervalued;
• a mixed or neutral classification means the AI signal is commercially irrelevant;
• AI recommendation behavior causes insurance purchasing behavior.

The current data measures the AI side of the relationship.

Financial usefulness must be established prospectively using the framework in Can AI Search Visibility Predict Revenue Growth? and the backtesting rules in How Investors Could Backtest AI Search Signals.

Methodology

The insurance-sector table is drawn from the same V0 public-parent panel used throughout the AI Investor Signals series.

Eligible company-level observations

The primary company data comes from structured company arrays that distinguish:

• company presence;
• valid recommendation status;
• recommendation rank;
• sentiment and framing fields;
• platform family;
• normalized prompt;
• source vertical;
• month;
• explicit not-mentioned records.

Explicit not-mentioned records are important because they provide a real denominator for coverage.

Matched-panel construction

The primary change compares the same normalized prompt on the same AI platform family in the base month and September 2026.

July is the base month for nine of the ten sector rows. Travelers uses August because July is unavailable in its matched panel.

Recommendation coverage

Recommendation coverage = valid recommendation cells / eligible matched prompt-platform cells

Change is reported in percentage points.

Failure handling

Explicit extraction failures are excluded. A failed extraction is treated as unknown, not as a legitimate zero recommendation.

Duplicate handling

Response-identical cross-vertical exports are collapsed in the primary cleaned panel so the same response state does not receive extra weight merely because it was stored in multiple overlapping datasets.

Parent-company mapping

Known consumer brands and entity variants are rolled to public parents. A parent is treated as recommended in a prompt-platform cell when at least one tracked entity mapped to that parent is recommended.

Exploratory intervals

The primary point estimate is cell-weighted. For uncertainty, matched-cell changes are averaged within normalized prompts, the standard error is calculated from the prompt-level means, and a normal 1.96 multiplier is applied around the point estimate.

These are exploratory measurement intervals, not causal confidence intervals.

V0 candidate rules

A positive AI divergence candidate requires:

• recommendation change of at least +5 percentage points;
• exploratory 95% interval lower bound above zero;
• at least four improving platforms.

A negative AI divergence candidate requires:

• recommendation change of -5 percentage points or worse;
• exploratory 95% interval upper bound below zero;
• at least four worsening platforms.

Everything else is Mixed / neutral.

The full methodology is documented in How We Measure AI Commercial Momentum.

Limitations

1. The panel is small. Ten mapped public parents are not representative of the full listed insurance industry.
2. The slice mixes insurance categories. Life, P&C, pet insurance, managed care, pharmacy-benefit exposure, and diversified financial services have different economics.
3. Travelers uses an August base month. Its change window differs from the July-to-September rows.
4. Prompts are research panels, not market-share-weighted consumer demand. Prompt frequency does not equal search volume, quote volume, premium, enrollment, or revenue.
5. Parent mapping is imperfect. Consumer-facing brands may represent only part of the public parent's economics.
6. AI platform behavior can change. A three-month period should not be treated as a permanent platform rule.
7. Recommendation coverage has not been linked to insurance purchases. The commercial pathway remains a hypothesis.
8. Cross-platform agreement is not financial validation. Platform breadth may improve signal quality, but that must be tested.
9. Absolute coverage and momentum are different. A company can have high coverage while declining, or low coverage while improving.
10. The current analysis does not normalize for product-market share. Competitive recommendation share and real-world market share require separate sector-specific denominator work.

What We Will Test Next

The insurance sector provides several natural prospective tests.

For consumer-facing insurance categories, potential downstream outcomes include:

• branded search movement;
• quote-start activity where available;
• website or app engagement;
• policy application activity;
• direct written premium growth in matched insurance lines;
• policy count or enrollment growth;
• analyst revenue or premium estimate revisions;
• reported segment revenue;
• membership or covered lives for health-plan businesses;
• revenue surprise and EPS surprise;
• later sector-relative stock performance after the commercial relationship is validated.

The strongest near-term research question is not whether the companies with positive AI momentum outperform as stocks.

It is whether the AI-side movement precedes a measurable change in the commercial behavior represented by the prompts.

If it does, the signal can advance through the maturity ladder described in What Is an AI Investor Signal?.

If it does not, the proper conclusion is that the AI recommendation change was a visibility measurement rather than a validated commercial leading indicator.

Related LLM Authority Index Research

Core framework

• Can AI Search Signal Future Revenue Growth? The AI Commercial Momentum Hypothesis
• Initial Findings From 25 Public Companies
• How We Measure AI Commercial Momentum
• AI Investor Signal Tracker
• AI Recommendation Share vs. Market Share
• Cross-Platform AI Visibility and Recommendation Portability

Insurance and health-plan company studies

• MetLife
• Principal Financial Group
• Molina Healthcare
• Cigna
• UnitedHealth Group / UnitedHealthcare
• Trupanion
• Travelers
• Allstate
• Lincoln Financial
• Corebridge Financial

Adjacent research

• Healthcare Stocks and AI Search
• Persistence-Portability Gap
• The Most-Cited Websites in AI for High-Stakes Insurance Consumer Decisions

Research Disclosure

LLM Authority Index is affiliated with CiteWorks Studio and related AI-search research and services. Commercial relationships do not change the frozen source data, matching logic, parent mapping, candidate thresholds, uncertainty rules, or published results.

No company can purchase a position, score, recommendation, or classification in this research.

The AI Investor Signals program is exploratory research. It is not investment advice, and the current AI recommendation measurements have not been validated as predictors of revenue, earnings, valuation, or stock returns.