Do AI Platforms Cite the Same Websites for High-Stakes Consumer Decisions? Comparing ChatGPT, Gemini, Perplexity, Copilot and Google AI Search

LLM Authority Index Research | High-Stakes Consumer Decisions | July-September 2026

AI platforms do not appear to share one universal citation ecosystem.

LLM Authority Index compared the 100 most frequently cited domains by response-level presence across six AI platform families using 60,881 deduplicated AI responses collected from July through September 2026 across 53 high-stakes consumer vertical datasets.

The six Top 100 lists contained 246 distinct domains. Of those, 117 domains, or 47.6%, appeared in only one platform's Top 100. Another 37 appeared in only two.

Only 15 domains appeared in the Top 100 for all six platforms.

The closest pair was Google AI Overviews and Google AI Mode, which shared 82 of their Top 100 domains. At the other end, Google AI Overviews and Microsoft Copilot shared only 35.

The median pairwise Jaccard similarity across the 15 platform pairs was 36.1%.

For publishers, CMOs, PR teams and agencies, the practical implication is straightforward: strong citation visibility on one AI platform should not be treated as proof of equivalent visibility on another.

This study is part of the LLM Authority Index research program on AI citations in high-stakes consumer decisions.

Answer Capsule

Do ChatGPT, Google AI Overviews, Google AI Mode, Gemini, Perplexity and Microsoft Copilot cite the same websites when consumers research high-stakes financial, insurance and medical decisions? Not consistently. Across the six platform-specific Top 100 lists in the LLM Authority Index July-September 2026 panel, there were 246 distinct domains. Only 15 domains ranked in the Top 100 on all six platforms, while 117 appeared in only one platform's Top 100. Google AI Overviews and Google AI Mode were the most similar pair, sharing 82 of 100 domains, while Google AI Overviews and Microsoft Copilot shared only 35. The median pairwise Top 100 Jaccard similarity was 36.1%.

Key Findings

• The analysis used 60,881 deduplicated AI responses across six public platform families.
• The underlying structurally cleaned corpus contained 278,499 citation events before public source-comparison infrastructure exclusions.
• For the cross-platform source comparison, obvious navigation and infrastructure artifacts were excluded, leaving approximately 275,500 citation events available for source-ecosystem analysis.
• The union of all six platform Top 100 lists contained 246 distinct registrable domains.
• 117 of 246 domains, or 47.6%, appeared in only one platform's Top 100.
• 153 of 246 domains, or 62.2%, appeared in no more than two platform Top 100 lists.
• Only 15 domains, or 6.1% of the Top 100 union, appeared in all six platform Top 100 lists.
• Google AI Overviews and Google AI Mode shared 82 of 100 domains, the highest overlap in the study.
• Google AI Overviews and Microsoft Copilot shared 35 of 100 domains, the lowest overlap in the study.
• The median pairwise Top 100 Jaccard similarity was 36.1%.
• Microsoft Copilot had 39 domains that appeared in its Top 100 and no other platform's Top 100, the largest platform-exclusive Top 100 count in this panel.
• ChatGPT had 28 platform-exclusive Top 100 domains, Gemini had 23, Perplexity had 15, Google AI Mode had 7 and Google AI Overviews had 5.

Why This Research Belongs in Daily AI Search Marketing Decisions

This study is directly relevant to day-to-day AI search marketing because publisher selection is a resource-allocation decision. If the source ecosystems of ChatGPT, Google AI Overviews, Google AI Mode, Gemini, Perplexity and Microsoft Copilot differ materially, then one universal outreach list will systematically miss publishers that matter on particular platforms.

The final six platform Top 100 lists reinforce that point. Their combined union contains 246 distinct domains. Only 15 domains appear in all six Top 100 lists, while 117 appear in only one. The median pairwise Jaccard similarity is 36.1%.

This study is the portability side of the Persistence-Portability Gap. Portability asks how consistently a leading citation source set transfers across AI platforms. Article 04 measures the related time dimension, while Article 01 combines both into the broader framework. Here, portability is a symmetric source-set comparison, not a probability that a publisher placement will transfer from one engine to another.

Commercial Action Matrix: What Publishers and CMOs Can Do With the Findings

The operational takeaway is simple: AI search teams should maintain a cross-platform core list, platform-specific source lists and category-specific source lists at the same time. The data should be used to decide where to investigate, pitch, partner, publish and monitor. It should not be used to promise that a citation or placement will cause a brand recommendation.

Questions This Study Answers

• Should a CMO use the same publisher outreach list for Google AI Overviews, ChatGPT and Perplexity?
• Which publishers have citation visibility across all major AI platforms rather than only one engine?
• Does a website that performs well in ChatGPT also tend to perform well in Google AI Overviews?
• Which AI platforms have the most similar citation-source ecosystems?
• Which platform has the most distinctive Top 100 source list in this high-stakes consumer dataset?
• Can a publisher credibly show advertisers that its AI citation visibility is cross-platform?
• Should brands build separate earned-media strategies for Google AI Overviews, Google AI Mode and ChatGPT?
• Which domains are repeatedly cited across credit, lending, insurance, investing, retirement, mortgage, health and medical decisions?
• How much of AI source selection appears platform-specific even among the most frequently cited websites?

How This Study Fits the High-Stakes Consumer Decisions Research Corpus

This article is the primary cross-platform source-overlap study in the LLM Authority Index High-Stakes Consumer Decisions research series.

• The 2026 AI Citation Authority Study establishes the overall source layer, category concentration and the combined persistence-versus-portability framework.
• The Persistence-Portability Gap measurement framework defines how temporal persistence and cross-platform source-set portability should be reported separately.
• How Stable Are AI Citations? measures the time dimension and same-prompt citation drift.
• Reddit's Decline in AI Citations shows how one major source can move materially even while the broader publisher portfolio remains persistent.
• Articles 05 through 08 break source concentration, persistence and portability into the four high-stakes decision families.
• Articles 09 through 14 provide the six platform-specific Top 100 source lists used to validate this article's final overlap calculations.

This division matters commercially. Article 03 answers where citation authority transfers across platforms. Article 04 answers how stable that authority is over time. The category and platform studies answer where a publisher is actually strongest. Those questions should be used together when building a publisher outreach or AI visibility program.

What We Compared

The comparison uses the same frozen July-September 2026 corpus underlying the 2026 AI Citation Authority Study.

The six public platform families are:

1. Google AI Overviews
2. Google AI Mode
3. ChatGPT
4. Gemini
5. Perplexity
6. Microsoft Copilot

The original source archive preserves separate Google keyword and non-keyword collection lanes. Those source surfaces are consolidated into the two public Google platform families only for editorial reporting.

Platform sample sizes

These totals should not be interpreted as platform market share or as evidence that one platform is inherently more important. Different platforms and source surfaces return different numbers of citations per response. The principal comparison therefore uses within-platform response-level domain presence rather than raw citation volume across engines.

High-Stakes Consumer Verticals Covered

The source corpus spans 53 vertical labels across four high-stakes consumer decision families.

Credit, Debt, Banking and Lending

Auto Refinance Loans; Bad Credit Loans; Best Banks; Certificates of Deposits; Credit Cards; Credit Cards for Building Credit; Credit Monitoring; Credit Repair; Debt Relief & Consolidation; Home Equity Loans; Money Market Accounts; Personal Loans and Online Lenders; Savings Account; Student Loan Refinance; Student Loans; Tax Relief.

Insurance

Car Insurance; Dental Insurance; Disability Insurance; Health Insurance; Life Insurance Companies; Long-Term Care Insurance; Medicare Supplement Insurance; Pet Insurance; Renters Insurance; Short Term Health Insurance; Travel Insurance; Vision Insurance.

Investing, Retirement, Mortgage and Financial Decisions

Annuities; Crypto Exchanges; Crypto Wallets; Gold IRAs and Precious Metals Dealers; Mortgage; Mortgage Refinance Lenders; Online Financial Advisors; Online Stock Brokers; Reverse Mortgage; Robo-Advisors; Structured Settlements.

Health and Medical

Addiction Treatment Centers; Assisted Living Facilities; Dental Implants; Fertility Clinics; Hearing Aids; Home Health Care; IVF Clinics; Medical Alert Systems; Mental Health Treatment Centers; Online Doctors; Online Pharmacies; Online Therapy; STD Tests; Weight Loss and Metabolic Health.

Closely related verticals can share prompts, and some delivered datasets contain adjacent questions. The corpus-wide methodology therefore includes explicit repeated-record handling rather than assuming every filename is an independent prompt universe.

The Six Platforms Share Only 15 Top 100 Domains

The following registrable domains appeared in the Top 100 for every one of the six platform families:

1. amazon.com
2. bankrate.com
3. businessinsider.com
4. credible.com
5. experian.com
6. finder.com
7. forbes.com
8. goodrx.com
9. healthline.com
10. hearingtracker.com
11. lendingtree.com
12. money.com
13. seniorliving.org
14. squaremouth.com
15. stockbrokers.com

This is the most conservative cross-platform group in the study: domains that were not merely cited somewhere by every engine, but ranked among the 100 most frequently present sources within every platform's own response set.

That distinction matters. A domain can technically appear on all six platforms while still being too infrequent to make one platform's Top 100.

This section answers

• Which publishers have the strongest evidence of broad cross-platform AI citation visibility?
• Which domains can document Top 100 citation visibility across Google, ChatGPT, Gemini, Perplexity and Copilot?
• Is a domain's overall AI citation rank enough to prove cross-platform strength?

The answer to the last question is no. Cross-platform breadth needs to be measured directly.

Nearly Half of the Combined Top 100 Universe Is Platform-Exclusive

The six Top 100 lists produce a union of 246 distinct domains.

The commercially important number is not just the 15 universal domains.

Nearly two-thirds of the combined Top 100 universe appears in no more than two platform lists.

That is why a single aggregated "AI publisher list" can conceal meaningful platform differences.

Which AI Platforms Cite the Most Similar Top Websites?

Because every list contains exactly 100 domains, there are two intuitive ways to compare them:

• Shared domains: how many domains appear in both Top 100 lists.
• Jaccard similarity: shared domains divided by the total unique domains across the two lists.

Pairwise Top 100 overlap

Median pairwise Jaccard similarity: 36.1%.

The two Google AI surfaces are clearly the closest pair in this study, but even they do not produce identical source ecosystems. Eighteen domains in one Top 100 list are absent from the other's Top 100.

Microsoft Copilot is the most separated platform by several measures in this particular panel. It shares only 35 Top 100 domains with Google AI Overviews, 37 with Gemini and 38 with ChatGPT.

That does not mean Copilot is "better" or "worse." It means its visible source selection was materially different in this measured high-stakes consumer corpus.

The Most-Cited Websites Differ Substantially by Platform

The divergence is visible even at the top of each ranking.

Google AI Overviews: Top 5 by response citation coverage

Google AI Mode: Top 5

ChatGPT: Top 5

Gemini: Top 5

Perplexity: Top 5

Microsoft Copilot: Top 5

The percentages above are within-platform response citation coverage. They should not be compared as if every platform had the same sample size, retrieval behavior or citation format.

They are useful for showing how the source hierarchy changes inside each platform.

A Publisher Can Be Dominant on One Platform and Minor on Another

Several large domains make the platform effect especially visible.

These results should be interpreted descriptively. A zero or near-zero value in this table means the domain was rarely or never captured as a visible citation in that platform's responses within this study. It does not prove the platform never uses or knows the source.

The pattern is still commercially meaningful. The same publisher can occupy very different positions in different AI source ecosystems.

Which Platforms Have the Most Platform-Specific Top 100 Sources?

A domain is counted as platform-exclusive here only if it appears in one platform's Top 100 and in none of the other five Top 100 lists.

This is not a measure of total web exclusivity. A domain outside another platform's Top 100 may still receive citations there.

It is a measure of how much each platform's high-frequency source tier differs from the other platforms' high-frequency source tiers.

Examples of platform-exclusive Top 100 domains in this panel

Google AI Overviews: Sallie Mae, Earnest, eHealthInsurance and SingleCare.

Google AI Mode: Anthem, Pets Best, International Insurance, ASPCA Pet Health Insurance, Instagram, Delta Dental and Insurance Business.

ChatGPT: Reuters, TechRadar, J.D. Power, Tom's Guide, AP News, Barron's, FDA.gov, GEICO and several other domains.

Gemini: ElderLife Financial, GreenFi, AutoInsurance.com, Marcus, Labcorp, MetaMask and Morningstar among others.

Perplexity: The New York Times, Verywell Mind, Koinly, Which?, CoinGecko, Healthcare.gov and several crypto and finance domains.

Microsoft Copilot: Wealthvieu, ClearValue Lending, BrokerChooser, Retirement Living, ConsumersAdvocate.org, NewMouth and several other niche consumer sites.

Because these examples come from the platform-specific Top 100 threshold, they should not be interpreted as claims that the domain is absent from every other engine.

What This Means for Publisher Sales Teams

A publisher seeking to demonstrate value to advertisers should separate overall AI visibility from platform-specific AI visibility.

Potential evidence in a publisher media kit could include:

• Top 100 status on one or more AI platforms;
• citation coverage on Google AI Overviews;
• citation coverage on Google AI Mode;
• citation coverage on ChatGPT;
• number of AI platforms where the publisher ranks in the Top 100;
• high-stakes consumer categories where the publisher is visibly cited;
• persistence of citation visibility across months.

A publisher appearing in all six platform Top 100 lists has a different commercial story from a publisher that is highly concentrated in one platform.

Neither position is automatically better. The useful question is whether the publisher's measured citation footprint matches the advertiser's target platform and consumer category.

This section answers

• Can publishers use independent AI citation data in advertiser sales materials?
• What evidence can a publisher show a brand that cares specifically about Google AI Overviews?
• How should a publisher describe strong ChatGPT visibility without implying guaranteed brand recommendations?

The safest framing is descriptive: measured citation presence, platform breadth, category breadth and persistence.

What This Means for CMOs and Brand PR Teams

A single universal publisher list is unlikely to be enough for brands that care about multiple AI platforms.

A more defensible workflow is:

1. Start with the cross-platform core. Identify publishers that repeatedly appear across several engines.
2. Add platform-specific targets. Build separate lists for Google AI Overviews, Google AI Mode and ChatGPT first if those are the highest-priority surfaces.
3. Filter by category. A publisher's overall AI prominence does not necessarily mean it is important in insurance, mortgage, credit repair or health.
4. Check persistence. Prefer evidence that persists across months rather than relying on a one-day snapshot.
5. Separate citation from recommendation. Use citation visibility to inform outreach, not as proof that a placement will cause an AI recommendation.

LLM Authority Index is publishing dedicated studies for each platform:

• The 100 Most-Cited Websites in Google AI Overviews for High-Stakes Consumer Decisions
• The 100 Most-Cited Websites in Google AI Mode for High-Stakes Consumer Decisions
• The 100 Most-Cited Websites in ChatGPT for High-Stakes Consumer Decisions
• The 100 Most-Cited Websites in Perplexity for High-Stakes Consumer Decisions
• The 100 Most-Cited Websites in Gemini for High-Stakes Consumer Decisions
• The 100 Most-Cited Websites in Microsoft Copilot for High-Stakes Consumer Decisions

Google AI Overviews and Google AI Mode Are Similar, But Not Interchangeable

Google's two AI surfaces produced the closest Top 100 source lists in this study.

They shared 82 domains, corresponding to a 69.5% Jaccard similarity.

That is much higher than any other pair.

But 82 shared domains also means 18 of the Top 100 domains differ between the two Google surfaces.

For a CMO focused specifically on Google's AI surfaces, that is an important distinction. A publisher strategy optimized for Google AI Overviews may transfer better to Google AI Mode than to ChatGPT or Copilot, but the transfer is not complete.

The category-specific studies will test whether that relationship remains equally strong in credit, insurance, financial and health decisions.

ChatGPT Is Not a Reliable Proxy for Google AI Search

ChatGPT shared:

• 54 of 100 domains with Google AI Overviews;
• 53 of 100 with Google AI Mode.

The corresponding Jaccard similarities were 37.0% and 36.1%.

This means there is meaningful overlap, but nearly half of each Top 100 list is different even before examining rank position or category-specific behavior.

A brand that measures only ChatGPT and assumes the result represents Google AI search is therefore leaving a substantial part of the measured source ecosystem unobserved.

Prior Research Also Finds Significant Cross-Engine Citation Fragmentation

Independent studies reach the same directional conclusion: AI engines often cite materially different source sets. The percentages should not be compared as if they were the same metric because the studies use different prompt populations, ranking depths, engines and denominators.

Wellows: Same-Question Source Overlap Across 22.7 Million Citations

Wellows analyzed 22.7 million citations across 1.15 million questions spanning ChatGPT, Gemini, Perplexity, Google AI Overviews and Google AI Mode. On the 531,889 questions where all five engines cited at least one website, 79.6% of website-and-question pairings appeared on exactly one engine, while 0.31% appeared on all five. Google AI Overviews and Google AI Mode were the highest-overlap pair in that study.

Source: Wellows, "89% of What ChatGPT Cites, Perplexity Never Touches"

That is a same-question measurement. Wellows asks: when the engines answer the same question, do they cite the same websites?

This article asks a different but commercially complementary question: across thousands of high-stakes consumer responses, how similar are the engines' recurring Top 100 source ecosystems?

Foglift: Buyer-Intent Top Lists Are Also Engine-Specific

Foglift's Q2 2026 benchmark tested 75 brand-neutral buyer-intent prompts across 25 verticals and five engines, producing 375 responses and 1,119 distinct cited domains. Its five Top 25 source lists produced a union of 81 domains. 50 of 81, or 61.7%, appeared in only one engine's Top 25, and only one domain appeared in all five. Foglift also reports a cross-engine Jaccard benchmark of 0.18 across the full prompt study.

Source: Foglift, "AI Search Citation Benchmark: Q2 2026"

Foglift is especially relevant because its prompt population is explicitly buyer-intent oriented. Its result independently supports the commercial conclusion that a publisher list built for one engine should not be assumed to transfer cleanly to another.

Writesonic: Identical Prompts Produce Low Source Agreement

Writesonic analyzed 161,286 prompts across ChatGPT, Gemini, Perplexity and Google AI Overviews. Among the 70,879 prompts where all four returned citations, Writesonic reported that only 3.8% of cited sources were shared across all four platforms. Pairwise domain-level Jaccard similarity ranged from 0.119 to 0.237.

Source: Writesonic, "Do AI Engines Cite the Same Sources? We Studied 161,286 Prompts Across 4 Platforms"

Again, the unit differs from this Top 100 ecosystem study. Writesonic measures same-prompt overlap. LLM Authority Index measures overlap among the most frequently recurring domains in a defined high-stakes corpus.

Machine Relations: Source Selection Patterns Differ by Engine

Machine Relations has also published six-engine research showing materially different source-selection patterns across ChatGPT, Perplexity, Gemini, Claude and Google's AI surfaces.

Source: Machine Relations, "How Six AI Engines Choose Sources"

The value of comparing these studies is not that they produce one universal overlap percentage. They do not. The value is that independent methods repeatedly show that AI source authority is engine-specific enough that single-engine measurement leaves meaningful blind spots.

Why Our Overlap Percentages Are Higher Than Some Same-Prompt Studies

A reader familiar with other AI citation studies may notice that LLM Authority Index reports a median 36.1% Top 100 Jaccard similarity, while same-prompt citation-overlap studies often report much lower percentages.

That does not mean one of the studies is wrong.

They are measuring different things.

Same-prompt overlap asks:

When two AI engines answer the exact same question, how many of the sources in those two individual answers match?

This can be extremely low because each answer may cite only a handful of sources.

Top 100 ecosystem overlap asks:

Across thousands of high-stakes commercial responses, how much do the two engines' most frequently recurring source domains overlap?

Large publishers can appear repeatedly in both ecosystems even if the two engines do not cite the same publisher on the exact same prompt.

For commercial planning, both measurements matter.

• Same-prompt overlap measures answer-level source agreement.
• Top-list overlap measures whether the same publishers repeatedly occupy the high-frequency source tier across platforms.

This article focuses on the second question.

How This Study Is Different

This study does not claim to be the first cross-engine citation-overlap analysis.

Its contribution is the combination of several design choices:

High-Stakes Consumer Decision Focus

The prompt universe concentrates on consequential consumer decisions involving credit, debt, lending, insurance, investing, retirement, mortgage, health and medical care rather than a broad all-topic sample.

Six Public Platform Families

The comparison includes both Google AI Overviews and Google AI Mode alongside ChatGPT, Gemini, Perplexity and Microsoft Copilot.

53 Source Vertical Labels

The source archive spans dozens of commercially valuable consumer verticals, allowing later category-specific overlap studies without requiring hundreds of thin individual-vertical articles.

Three-Month Panel

The July-September structure allows the project to distinguish persistent source patterns from short-lived citation spikes.

Response-Level Ranking

A domain counts no more than once per response for the primary ranking, reducing the effect of an answer that links repeatedly to the same website.

Registrable-Domain Consolidation

Subdomains are consolidated using the Public Suffix List before cross-platform comparison.

Explicit Failure and Repeat Handling

Records explicitly marked as extraction failures are excluded from valid-response analysis, and exact repeated records across overlapping source datasets are counted once in the corpus-wide analysis.

Methodology

Source Corpus

The frozen July-September source archive contains 68,184 raw observations and 310,071 raw citation entries before analytical cleaning.

The same anti-double-counting and extraction-failure rules used in the flagship study produce 60,881 deduplicated valid AI responses and 278,499 structurally retained citation events.

For this platform-source comparison, obvious navigation and infrastructure artifacts are removed from public source tables. Approximately 275,500 citation events remain after that additional source-quality filter.

Platform Families

The public platform family comes from the preserved source provenance. Google keyword and non-keyword collection lanes are grouped into Google AI Overviews and Google AI Mode for editorial comparison.

Domain Normalization

Citation URLs are parsed to hostnames and consolidated to the registrable domain using the Public Suffix List.

Examples:

• money.usnews.com becomes usnews.com;
• finance.yahoo.com becomes yahoo.com;
• pharmacy.amazon.com becomes amazon.com.

Primary Platform Ranking Metric

For each platform, each domain is counted at most once within a response.

For platform p and domain d:

platform response citation coverage = responses on platform p citing domain d / eligible deduplicated responses on platform p

The Top 100 for each platform consists of the 100 domains with the highest domain-response appearance counts on that platform.

Pairwise Top 100 Overlap

For two platform Top 100 sets A and B:

shared domains = |A ∩ B|

Jaccard similarity = |A ∩ B| / |A ∪ B|

Because each set contains 100 domains, the shared-domain count is especially easy to interpret. A shared count of 82 means 82 domains occur in both Top 100 lists.

Platform-Exclusive Top 100 Domain

A domain is called platform-exclusive in this article only if it appears in one platform's Top 100 and in none of the other five Top 100 lists.

This does not mean the other platforms never cited the domain.

Infrastructure Exclusions

Obvious non-editorial infrastructure such as Bing image-host assets, certain Bing navigation/map wrappers, OpenAI image hosts and Google account-activity pages are removed from public platform source comparisons.

Platform-owned substantive pages are not automatically removed merely because the platform owns the domain.

Missing Months

Several vertical datasets were never produced for July. They are treated as unavailable rather than zero-filled. Platform rankings therefore describe the entire measured July-September corpus rather than a perfectly balanced three-month panel.

Limitations

1. This study measures visible citations, not model training data or hidden retrieval steps.
2. Top 100 ecosystem overlap is not the same as same-prompt citation overlap. Readers should not compare the percentages directly without accounting for the metric.
3. Platform sample sizes differ. Response-level ranking reduces but does not eliminate differences in collection design and citation behavior.
4. The prompt universe is intentionally commercial and high-stakes. Results are not representative of every AI query.
5. Some source verticals overlap or contain adjacent prompts. Exact repeat handling reduces corpus-wide duplication, but source taxonomy remains an important caveat.
6. Citation extraction behavior can differ by platform. A platform with fewer visible source links can produce a different observable ecosystem from a platform that exposes many citations.
7. A Top 100 threshold is useful but arbitrary. A domain ranked #101 is not meaningfully different from #100 simply because of the cutoff.
8. Platform behavior changes. These findings describe the July-September 2026 panel.
9. Citation does not prove recommendation causality. A cited publisher may inform an answer without causing a company to be recommended.

Dataset and Downloads

Planned public research assets for this article include:

• six platform-specific Top 100 domain lists;
• the complete pairwise Top 100 overlap matrix;
• response citation coverage by platform;
• platform-exclusive Top 100 domain flags;
• cross-platform breadth counts;
• downloadable CSV and methodology notes.

Platform overlap CSV placeholder: {{PLATFORM_OVERLAP_CSV_URL}}

Canonical study URL: https://llmauthorityindex.com/resources/research/ai-platform-citation-overlap-high-stakes-consumer-decisions

Related LLM Authority Index Research

Foundation and measurement framework

• The 2026 AI Citation Authority Study
• The Persistence-Portability Gap
• Reddit's Decline in AI Citations Across High-Stakes Consumer Decisions
• How Stable Are AI Citations for High-Stakes Consumer Decisions?

Decision-family source studies

• The Most-Cited Websites in AI for High-Stakes Credit, Debt, Banking and Lending Consumer Decisions
• The Most-Cited Websites in AI for High-Stakes Insurance Consumer Decisions
• The Most-Cited Websites in AI for High-Stakes Investing, Retirement, Mortgage and Financial Consumer Decisions
• The Most-Cited Websites in AI for High-Stakes Health and Medical Consumer Decisions

Platform-specific Top 100 source studies

• The 100 Most-Cited Websites in Google AI Overviews for High-Stakes Consumer Decisions
• The 100 Most-Cited Websites in Google AI Mode for High-Stakes Consumer Decisions
• The 100 Most-Cited Websites in ChatGPT for High-Stakes Consumer Decisions
• The 100 Most-Cited Websites in Gemini for High-Stakes Consumer Decisions
• The 100 Most-Cited Websites in Perplexity for High-Stakes Consumer Decisions
• The 100 Most-Cited Websites in Microsoft Copilot for High-Stakes Consumer Decisions

These links follow the final published slug structure. The six platform-specific studies are the source lists used to reconcile the final overlap table in this article.

References

1. LLM Authority Index. The 2026 AI Citation Authority Study.
2. LLM Authority Index. The Persistence-Portability Gap.
3. Wellows. 89% of What ChatGPT Cites, Perplexity Never Touches.
4. Foglift Research. AI Search Citation Benchmark: Q2 2026.
5. Writesonic. Do AI Engines Cite the Same Sources? We Studied 161,286 Prompts Across 4 Platforms.
6. Machine Relations. How Six AI Engines Choose Sources.