The 100 Most-Cited Websites in Microsoft Copilot for High-Stakes Credit, Debt, Banking and Lending Consumer Decisions

LLM Authority Index Research | Microsoft Copilot | Credit, Debt, Banking and Lending | July-September 2026

Microsoft Copilot uses a finance-heavy source ecosystem when consumers ask high-stakes questions about credit, debt, banking and lending.

Across 2,379 deduplicated Microsoft Copilot responses covering 16 financial decision verticals, NerdWallet ranked #1, appearing as a cited domain in 23.33% of eligible responses. Forbes ranked #2 at 21.77%, Bankrate #3 at 20.85%, CNBC #4 at 20.13%, and U.S. News #5 at 14.80%.

The category-specific ranking is substantially more concentrated than Copilot overall. The Top 10 financial domains account for 48.8% of domain-response appearances, compared with 30.0% for Copilot across all four high-stakes decision families. The financial Top 100 account for 81.3%, compared with 56.7% in the broader Copilot study.

Only 47 of the financial Top 100 domains also appear in Copilot's overall Top 100. Bankrate is especially category-specific: its financial response coverage is 2.44 times its Copilot-wide rate. WalletHub rises to 2.97 times its platform-wide coverage.

Copilot also requires a source-audit step that is less visible in several other platforms. In the deduplicated financial slice, 705 captured citation entries pointed to Bing search, thumbnail, image, favicon/API, or OpenStreetMap navigation infrastructure. Those entries were excluded from the publisher ranking while substantive source links remained.

This is a source-visibility ranking, not a trust score, quality grade, recommendation score, or causal ranking. A domain can be frequently cited without causing Copilot to recommend a specific company or financial product.

Answer Capsule

Which websites does Microsoft Copilot cite most often when answering high-stakes credit, debt, banking and lending questions?

NerdWallet leads this measured Copilot financial panel with 23.33% response citation coverage, followed by Forbes at 21.77%, Bankrate at 20.85%, CNBC at 20.13%, and U.S. News at 14.80%. The analysis covers 2,379 deduplicated Copilot responses, 7,042 public-facing citation events, 767 registrable cited domains, and 6,427 domain-response appearances across 16 financial decision verticals.

Key Findings

• NerdWallet ranks #1 with 555 responses citing the domain, equal to 23.33% of the Copilot financial sample.
• Forbes ranks #2 at 21.77%, followed by Bankrate at 20.85%, CNBC at 20.13%, and U.S. News at 14.80%.
• The Copilot financial slice contains 2,379 eligible responses, 7,042 public-facing citation events, 767 cited domains, and 6,427 domain-response appearances.
• 705 technical or navigational entries were excluded from the public ranking: 451 from bing.com, 237 from bing.net, 14 from bingapis.com, and 3 from openstreetmap.org.
• The excluded technical layer represented 9.1% of captured citation entries in the deduplicated Copilot financial slice.
• The Top 10 account for 48.8% of domain-response appearances and 47.4% of citation events.
• The Top 100 account for 81.3% of domain-response appearances and 81.2% of citation events.
• Only 47 of the financial Top 100 also appear in Copilot's overall Top 100 across the full High-Stakes Consumer Decisions study.
• 46 domains appear in the Copilot financial Top 100 in July, August, and September, with a 39.9% median monthly Top 100 Jaccard similarity.
• Forbes ranks #1 in July, NerdWallet ranks #1 in August, and Bankrate ranks #1 in September.
• The August and September Top 100 lists are the most similar monthly pair, sharing 67 domains with a 50.4% Jaccard similarity.
• Copilot source leadership changes materially by financial vertical. Bankrate leads Auto Refinance and Home Equity Loans, CardRates leads Credit Cards, WalletHub leads Credit Cards for Building Credit, and U.S. News leads Student Loans and Student Loan Refinance.
• Copilot's financial source hierarchy also differs from the other AI engines in the parent category study, supporting platform-specific rather than universal publisher targeting.

Questions This Study Answers

• Which websites dominate Microsoft Copilot citations for high-stakes Credit, Debt, Banking and Lending decisions, and how does that source hierarchy change by vertical?
• How persistent is Copilot's financial source market over time, and how portable is that authority to other AI platforms?
• What should financial publishers, lenders, banks, credit brands and CMOs measure before allocating resources to Copilot-focused content, PR, partnerships or earned media?

Why This Research Belongs in Daily AI Search Marketing Decision

This article sits at the intersection of two commercially useful views of the same source market:

• the Microsoft Copilot platform view, which asks which sources recur across high-stakes consumer decisions on Copilot; and
• the Credit, Debt, Banking and Lending decision-family view, which asks which sources recur across six AI platform families for one commercial financial category.

For publishers and specialist financial sites, that intersection supports a more precise claim than saying a domain is simply "visible in Copilot." A source can document whether it is cited specifically for high-intent credit, debt, banking and lending decisions, whether that visibility persists over time, and whether its authority is broad across financial verticals or concentrated in one specialist decision journey.

For brands, lenders, banks, fintechs, CMOs, PR teams and agencies, the operating question is:

Which sources are already present around the exact financial decisions our customers ask Microsoft Copilot to help them make?

Six findings matter for daily planning:

• the Top 100 account for 81.3% of domain-response appearances in this Copilot financial slice;
• the Top 25 account for 64.4% of domain-response appearances;
• only 47 domains overlap between this financial Top 100 and Copilot's broader High-Stakes Top 100;
• 46 domains remained in the monthly financial Top 100 in July, August and September, and the median of the three monthly Top 100 Jaccard comparisons was 39.9%;
• the final category child-study Top 100 lists show Copilot sharing 48 domains with ChatGPT, 47 with Perplexity, 44 with Google AI Mode, 43 with Google AI Overviews and 36 with Gemini; and
• 705 Bing, Bing API, Bing network and OpenStreetMap utility entries were excluded before constructing the publisher leaderboard.

The last point is especially important for Copilot. Microsoft documents that web-grounded Copilot responses can generate queries that are sent to the Bing search service. A raw capture can therefore contain technical, search, image, map, navigation and other utility URLs that should not be interpreted as editorial publisher authority. Microsoft Learn: web search in Microsoft Copilot

The parent Credit, Debt, Banking and Lending Citation Authority Study provides the category-level Persistence-Portability Gap interpretation. It reports 77.0% temporal persistence, 44.9% cross-platform portability, and an approximate 32.1 percentage-point Persistence-Portability Gap for the category.

This article answers a narrower Copilot-specific question. Its 39.9% median monthly Top 100 persistence should not be subtracted from one pairwise platform overlap to create a Copilot-specific gap. The parent PPG uses category-level summary statistics across the broader family.

The concentration result also should not be confused with stability. Copilot's financial source market is concentrated, yet more than half of the monthly Top 100 membership can differ across some month pairs. Concentration, persistence and portability are distinct measurements.

Citation visibility should also remain separate from brand mentions, recommendation rate, ranking, sentiment, citation-recommendation coupling and downstream outcomes. A source can be frequently cited without causing Copilot to recommend a particular lender, bank, card issuer or financial company.

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

The daily operating model is therefore four-layered:

1. Current Copilot financial core: maintain the sources that matter now, with the collection date attached.
2. Vertical and prompt-cluster layer: identify which sources dominate the exact credit repair, credit card, banking, student loan, personal loan or debt decision.
3. Longitudinal check: monitor whether priority sources persist, decline, rotate or change rank.
4. Cross-platform check: determine whether the same sources also matter in Google AI Overviews, Google AI Mode, ChatGPT, Perplexity and Gemini.

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

This article is the intersection of two parent studies:

• The Microsoft Copilot Platform Study
• The Credit, Debt, Banking and Lending Citation Authority Study

It should also be read with:

• the 2026 AI Citation Authority Study;
• the Persistence-Portability Gap;
• Do AI Platforms Cite the Same Websites?;
• How Stable Are AI Citations?;
• Reddit's Decline in AI Citations for Credit, Debt, Banking and Lending Decisions; and
• How Reddit Citations Changed in Microsoft Copilot for Credit, Debt, Banking and Lending Decisions.

This study isolates Microsoft Copilot responses associated with the Credit, Debt, Banking and Lending family of the LLM Authority Index High-Stakes Consumer Decisions corpus.

The family contains 16 source vertical labels:

1. Auto Refinance Loans
2. Bad Credit Loans
3. Best Banks
4. Certificates of Deposits
5. Credit Cards
6. Credit Cards for Building Credit
7. Credit Monitoring
8. Credit Repair
9. Debt Relief & Consolidation
10. Home Equity Loans
11. Money Market Accounts
12. Personal Loans and Online Lenders
13. Savings Account
14. Student Loan Refinance
15. Student Loans
16. Tax Relief

The archived Copilot financial slice contained 2,600 raw July-September observations before failure handling. 14 observations explicitly marked as extraction failures were excluded rather than treated as zero-citation responses. After exact analytical repeats were removed, 2,379 eligible responses remained.

The primary ranking metric is Response Citation Coverage: the percentage of eligible Copilot responses that cited a domain at least once. A domain counts no more than once within a response for the primary metric, even when Copilot surfaces multiple pages from that domain.

This section answers the following questions

• What denominator and primary ranking metric does this study use?
• Which 16 financial verticals are represented?
• How many technical Bing/OpenStreetMap entries were removed, and why are those exclusions important?

The Top 25 Microsoft Copilot Financial Citation Sources

The top five are unusually close compared with several other platform-specific studies. NerdWallet, Forbes, Bankrate, and CNBC each appear in roughly one-fifth or more of eligible responses, while U.S. News appears in nearly 15%.

The first tier mixes broad business publishers with finance-specialist publishers. NerdWallet and Bankrate have category depth, while Forbes and CNBC bring broad business authority. U.S. News combines general publishing breadth with substantial financial coverage.

WalletHub ranks #6 at 9.63%, almost three times its Copilot-wide response coverage. ClearValueLending, LendEDU, AutoLoanRate, and CardRates also become much more important in the financial slice than in Copilot overall.

This section answers the following questions

• Who leads Copilot citations for financial consumer decisions?
• Are the leading sources broad publishers, financial specialists, or both?
• Which finance-specific domains become more important inside the category?

The Full Top 100 Microsoft Copilot Financial Citation Sources

The Top 100 shows a strong financial head and a much longer specialist tail. Beyond the obvious category leaders, the list contains credit unions, lenders, card publishers, student-loan publishers, credit-monitoring sources, bank-rate sites, and government or consumer-protection sources.

This section answers the following questions

• Which 100 domains had the greatest response-level citation coverage in the measured Copilot financial panel?
• How did each domain rank in July, August, and September?
• Which specialist sources appear below the broad category leaders?

Citation Concentration

Copilot's financial source market is substantially more concentrated than Copilot's overall High-Stakes Consumer Decisions source market.

The commercial implication is that Copilot financial citation visibility is more top-heavy than Copilot overall. The Top 25 financial domains account for almost two-thirds of measured response-level domain appearances. That gives brands a manageable high-priority publisher set, but vertical specialists still matter because leadership changes sharply by product category.

This section answers the following questions

• Is Copilot's financial source market more concentrated than Copilot overall?
• How much visibility is captured by the Top 10, Top 25, Top 50, and Top 100?
• Can a financial brand focus only on the largest publishers?

Financial Specialization Versus Copilot Overall

Only 47 domains in this financial Top 100 also appear in Copilot's overall Top 100 across all high-stakes decision families. More than half of the financial Top 100 would therefore be missed by a strategy based only on the broad Copilot leaderboard.

Several specialist sources show large category lift. Bankrate is about 2.44x its Copilot-wide coverage, WalletHub about 2.97x, AutoLoanRate about 3.30x, CardRates about 3.05x, and LendEDU about 2.90x.

Forbes behaves differently. Its 21.77% financial coverage is only modestly above its 20.51% Copilot-wide coverage, showing that its strength is broad rather than narrowly financial. NerdWallet also gains inside finance, rising from 15.29% overall to 23.33% in this category.

This section answers the following questions

• Which sources are disproportionately important inside financial decisions?
• Which high-ranking Copilot domains are broad authorities versus financial specialists?
• How much information is lost when marketers use only Copilot's overall Top 100?

The Most-Cited Sources by Individual Financial Vertical

The category-wide ranking should not be treated as the answer for every financial product. The leading sources change materially by vertical.

Several patterns stand out:

• Bankrate leads Auto Refinance Loans, Certificates of Deposits, Home Equity Loans, and Savings Account in this Copilot slice.
• NerdWallet leads Bad Credit Loans and Personal Loans and Online Lenders.
• CardRates leads Credit Cards, while WalletHub leads Credit Cards for Building Credit.
• Forbes leads Best Banks and Credit Monitoring.
• CNBC leads Credit Repair.
• U.S. News leads both Student Loan Refinance and Student Loans.
• Debt Relief & Consolidation and Tax Relief have much smaller Copilot samples than the major banking and lending verticals and should be interpreted cautiously.

Some rows also surface adjacent sources because the originating datasets can contain shared or neighboring prompts. These are descriptive source views, not claims that every vertical is statistically independent.

This section answers the following questions

• Which publishers lead specific financial product categories in Copilot?
• Should a credit card company use the same publisher list as a savings account, student loan, or credit repair company?
• Which verticals have sample sizes too small for strong standalone conclusions?

Breadth Versus Specialization

A category leader can earn its position through breadth, depth, or both.

NerdWallet appears across 15 of 16 financial verticals and ranks #1 overall. Forbes also appears across 15 of 16, while U.S. News appears across 15 and CNBC across 14. Bankrate appears across 14 verticals but has exceptionally high coverage inside categories such as Auto Refinance, Home Equity Loans, Money Market Accounts, and Savings Account.

Specialists can matter despite narrower breadth. AutoLoanRate appears in only three financial verticals but ranks #12 overall because of its strength in Auto Refinance. CardRates also appears in only three verticals but ranks #14 overall because it leads the Credit Cards slice. Security.org is visible in only one financial vertical, Credit Monitoring, yet ranks #25 category-wide because its coverage is concentrated there.

For publishers, breadth and specialization support different commercial stories. For brands, they support different outreach strategies. A multi-product bank may value broad source presence, while a single-product company may gain more from a narrow specialist with high category penetration.

This section answers the following questions

• Which sources have broad financial-category reach?
• Can a narrowly specialized publisher rank highly without broad category coverage?
• How should a multi-product bank think differently from a single-product lender?

Monthly Persistence and Citation Drift

The Copilot financial Top 100 changed materially across the three-month window, even though several leaders remained near the top.

46 domains appeared in the financial Top 100 in all three months. The median of the three monthly Top 100 Jaccard comparisons was 39.9%.

The leadership sequence changed. Forbes ranked #1 in July, NerdWallet #1 in August, and Bankrate #1 in September. CNBC remained #3 in all three months. U.S. News moved from #6 in July to #5 in August and #4 in September.

The monthly movement is not evidence of a specific algorithm change. It can reflect retrieval behavior, source changes, prompt composition, platform changes, or normal citation turnover. The separate LLM Authority Index citation-drift study is designed to examine same-prompt source changes more directly.

This section answers the following questions

• Did Copilot's financial source market reset each month?
• Which publishers maintained leadership and which changed position?
• Can stable category authority coexist with substantial monthly source turnover?

Platform Differences: Copilot's Financial Source Market Is Not Interchangeable With Other AI Engines

The final Credit, Debt, Banking and Lending child-study Top 100 lists show meaningful cross-platform overlap, but no other measured engine uses the same source portfolio.

Copilot's closest sibling source set in this category is ChatGPT, but even that pair shares fewer than half of its Top 100 domains. Gemini is the farthest of the five sibling engines.

The ranking order also differs materially. Copilot's leading five are NerdWallet, Forbes, Bankrate, CNBC and U.S. News. ChatGPT elevates Forbes and The Wall Street Journal. Gemini elevates Money.com, Experian and WalletHub. Perplexity places Money.com and LendingTree in its Top 5.

That difference is commercially important. A source strategy optimized for Gemini, Perplexity, ChatGPT or a Google AI surface is not automatically a Copilot source strategy.

The operating implication is to maintain three layers: a cross-platform financial core, a Copilot-specific source layer, and a vertical-specific specialist layer.

This section answers the following questions

• Do the same publishers lead every AI platform for financial decisions?
• Which engines share the most Copilot financial sources?
• Should a financial brand use one universal outreach list for all AI platforms?

How This Compares With Other Microsoft Copilot Citation Research

Ahrefs: broad U.S. all-topic Copilot queries

Ahrefs' September 2026 Copilot citation study analyzes more than 3 million U.S. queries spanning all topics and ranks domains by mention share among the top cited sources. Its leaders are Amazon at 16.8%, Walmart at 12.6%, Wikipedia at 7.6%, ScienceInsights at 3.5% and EngineerFix at 3.3%. Ahrefs: The 50 Most-Cited Websites in Copilot

That is very different from this financial decision panel, where NerdWallet, Forbes, Bankrate, CNBC and U.S. News lead by response citation coverage.

The difference is not a contradiction. The prompt population, category mix, ranking metric, denominator and time-window definitions differ.

Tinuiti: commercial category tracking across six AI platforms

Tinuiti's Q3 2026 AI Citation Trends Report tracks ChatGPT, Perplexity, Google AI Mode, Google AI Overviews, Gemini and Microsoft Copilot across multiple commercial categories. Its current public summary emphasizes that citation behavior continues to differ by platform and category. Tinuiti Q3 2026 AI Citation Trends Report

That supports the need to measure Copilot separately and to define the commercial query population before interpreting a citation ranking.

Microsoft: Copilot web grounding uses Bing search

Microsoft's current documentation says that when web search is enabled, Microsoft Copilot and Copilot Chat can generate search queries derived from the user's prompt and send those queries to the Bing search service. Microsoft also documents web-search query citations in Copilot Chat. Microsoft Learn: Data, privacy, and security for web search in Microsoft Copilot

That architecture makes technical-source auditing especially important for Copilot datasets. Article 33 excludes Bing search, thumbnail, network, API and navigation infrastructure from the publisher ranking rather than treating utility URLs as editorial authority.

Machine Relations: broader multi-engine concentration uses a different denominator

Machine Relations' September 2026 open-web concentration release covers 22,179 cited domains, 15,782 observed answer runs and 110,877 domain-run citations across its 125-day observation window. Its methods note explicitly cautions that concentration measured on domain-run citations is not the same as concentration measured on citation events or other denominators. Machine Relations: How Concentrated Are AI Citations?

That is directly relevant to Article 33's interpretation. Copilot finance is highly concentrated under response-level domain appearances, but that percentage should not be merged with a multi-engine domain-run concentration statistic.

Studies can disagree because of:

• query selection
• commercial intent
• geography
• time window
• ranking metric
• citation-event versus response-level counting
• source-type classification
• technical-asset exclusions
• domain normalization
• platform mode and UI behavior

The relevant question is not which ranking is universally correct. It is which ranking matches the decision environment and measurement contract a publisher or brand actually cares about.

How This Study Is Different

This study is differentiated by the intersection of one platform, one commercially consequential family, and one frozen multi-month dataset.

• It measures only Microsoft Copilot inside a defined Credit, Debt, Banking and Lending decision family.
• It uses response citation coverage as the primary ranking metric rather than raw URL counts or share of only a preselected top-source set.
• It publishes all 16 financial vertical cuts inside the same substantive study instead of expanding the project into hundreds of thin pages.
• It separates captured technical Bing/OpenStreetMap infrastructure from substantive publisher citations.
• It reports July, August, and September ranks so aggregate authority can be distinguished from persistence.
• It compares the financial leaderboard with Copilot overall and with the other five AI platform families measured in the same parent corpus.
• It treats citation visibility as an observable relationship, not proof that a publisher caused a recommendation.

Why AI Citation Studies Can Disagree

Two well-run Copilot citation studies can produce different rankings without either one being wrong.

• Prompt universe: all-topic prompts, enterprise prompts, shopping prompts, or high-stakes financial decisions surface different source markets.
• Metric: response citation coverage, raw citation events, mention share, cited pages, or share among a Top 50 answer different questions.
• Platform surface: consumer Copilot, Copilot Chat, embedded Microsoft 365 experiences, and other product contexts do not necessarily expose citations identically.
• Technical cleanup: counting Bing thumbnails, image hosts, search wrappers, map directions, or favicons can materially change raw link totals.
• Domain normalization: subdomains such as money.usnews.com can be counted separately or consolidated to usnews.com.
• Time period: source selection can change across months even when a publisher remains strong in aggregate.
• Geography and personalization: studies can differ in market, session state, enterprise controls, or web-search availability.
• Deduplication and failure handling: overlapping exports, extraction failures, and repeated prompt records can change denominators if treated differently.

The correct question is therefore not, "Which study has the universal Copilot citation ranking?" The better question is, "Which methodology matches the decision population and measurement goal I care about?"

What This Means for Financial Publishers

Publishers should separate broad Copilot authority from financial Copilot authority. The category data gives publishers a defensible way to describe where they are actually visible.

• A broad publisher such as Forbes can point to both platform-wide and financial category visibility, but should preserve the study period and measured population.
• A finance specialist can document category lift when its financial response coverage materially exceeds its Copilot-wide rate.
• Publishers should report vertical breadth separately from category rank. Broad presence across many advertiser categories is a different commercial asset from deep leadership in one product category.
• Multi-month persistence is stronger evidence than a one-month spike. A publisher that stays near the top across July, August, and September has a stronger persistence story.
• Publishers should not claim that citation rank proves Copilot recommends their advertisers or that buying coverage will create future citations.

This section answers the following questions

• How can a financial publisher responsibly use Copilot citation data in sales or PR?
• What is the difference between broad platform authority and category authority?
• Which claims should publishers avoid making from citation rankings?

What This Means for CMOs and Financial Brands

For a financial brand, the useful output is not a generic list of websites that "AI likes." It is a prioritized source map for the exact platform and decision category.

• Use the category Top 25 as a high-priority research set, then layer the brand's own product vertical on top of it.
• Do not assume the best source targets for ChatGPT, Gemini, or Perplexity are the best source targets for Copilot.
• Separate broad publishers from specialists. A high-lift specialist may be more relevant to one product than a larger general publisher.
• Use persistence and vertical breadth as prioritization signals, but do not treat citation visibility as guaranteed recommendation influence.
• Monitor technical capture quality. Copilot produces a meaningful layer of Bing and navigation infrastructure that should not be mistaken for publisher authority.
• Repeat the measurement over time. The July-to-September Top 100 overlap shows that the source market moves enough for one-time audits to age.

This section answers the following questions

• How should a CMO turn the Copilot source map into a publisher-prioritization workflow?
• Should Copilot outreach targets be copied from another AI engine?
• How should brands use category lift, breadth, and persistence without overclaiming causality?

Methodology

Research source

This study uses the frozen July-September 2026 LLM Authority Index High-Stakes Consumer Decisions archive used by the broader 2026 AI Citation Authority Study.

Platform filter

Only observations assigned to the Microsoft Copilot platform family are included. Copilot remains separate from ChatGPT, Gemini, Perplexity, Google AI Overviews, and Google AI Mode.

Category assignment

Observations are included when their preserved source membership belongs to the Credit, Debt, Banking and Lending family. The 16 source verticals are listed in the What We Studied section.

Exact file deduplication

Byte-identical source files were deduplicated by SHA-256 before observation processing.

Analytical anti-double-counting

Within the category-platform slice, observations were treated as exact analytical repeats when they shared the same report month, exact raw platform/source surface, normalized prompt text, and exact set of original citation URLs. One analytical record was retained while preserving originating vertical memberships.

Prompt normalization

Prompt matching uses Unicode NFKC normalization, case folding, whitespace collapse, and trimming. Prompt wording is otherwise preserved.

Explicit extraction failures

The raw Copilot financial archive contained 14 observations explicitly marked as extraction failures. Those records were excluded from valid-response denominators rather than counted as zero-citation answers.

Domain normalization

Citation domains were normalized to lowercase, leading www. was removed, and subdomains were consolidated to registrable domains using Public Suffix List logic. For example, money.usnews.com is reported under usnews.com.

Copilot capture audit and technical exclusions

The deduplicated financial response set contained 7,747 captured citation entries before public infrastructure filtering. The public publisher ranking excludes 705 obvious technical or navigational entries, consisting of 451 bing.com, 237 bing.net, 14 bingapis.com, and 3 openstreetmap.org entries.

These include Bing maps/directions, search wrappers, image and thumbnail infrastructure, favicon/API assets, and OpenStreetMap navigation or copyright utility URLs. The exclusion is based on the role of the surfaced URL, not simply on Microsoft ownership. Substantive publisher pages remain eligible.

Primary metric

Response Citation Coverage = eligible Copilot financial responses citing domain d at least once / all eligible Copilot financial responses. The denominator for this article is 2,379 responses. A domain counts no more than once per response for this metric.

Secondary metric

Citation events are retained as a secondary metric. A response can contain multiple citation events from the same domain even though the domain counts once for response coverage.

Monthly ranks

July, August, and September Top 100 lists are calculated independently from eligible responses in each month using the same public-source rules.

Vertical views

A deduplicated response can retain more than one originating source membership when overlapping exports contain the same exact observation. Vertical tables are therefore descriptive source views and are not assumed to be statistically independent.

Jaccard overlap

Monthly Top 100 Jaccard similarity is calculated as shared Top 100 domains / unique domains across the two Top 100 lists.

No causal interpretation

A visible citation establishes an observed source relationship. It does not prove that the citation caused Copilot to mention, rank, recommend, or convert a consumer for any company.

Limitations

1. This is a defined high-stakes financial prompt corpus, not every Microsoft Copilot query about finance.
2. The 16 source vertical labels are not statistically independent. Some prompt banks overlap or contain adjacent questions.
3. Sample sizes differ materially by vertical. Debt Relief & Consolidation and Tax Relief are much smaller than the largest banking and lending cuts.
4. Copilot is not one immutable product surface. Product context, web-search settings, enterprise controls, and user experience can differ.
5. Visible citations are not hidden retrieval or training data. A source can influence an answer without appearing as a visible citation.
6. Citation presence is not recommendation causality. A cited domain does not prove a company on that domain will be recommended.
7. The Copilot capture stream includes a substantial technical layer. The published source ranking depends on transparent infrastructure filtering.
8. Registrable-domain aggregation hides page-level differences. The same domain can contain multiple content types and commercial intents.
9. The study window covers July through September 2026. Copilot source behavior can change after the endpoint.
10. Cross-study percentages should not be compared without aligning prompts, dates, denominators, metrics, technical filtering, and platform surfaces.
11. The study is observational. It measures what was cited, not why Copilot selected a source.

Commercial Disclosure

LLM Authority Index is part of a business ecosystem that provides AI visibility, research, and marketing services. Affiliated businesses may have current or historical commercial relationships with companies or publishers that appear in the dataset.

Commercial relationships are not inputs to the ranking methodology. Domains are included based on observed citation data. No ranking is an endorsement, and no commercial relationship is presented as the cause of a citation or recommendation outcome.

References

1. LLM Authority Index: The Most-Cited Websites in AI for High-Stakes Credit, Debt, Banking and Lending Consumer Decisions
2. LLM Authority Index: The 100 Most-Cited Websites in Microsoft Copilot for High-Stakes Consumer Decisions
3. LLM Authority Index: The Persistence-Portability Gap
4. LLM Authority Index: Do AI Platforms Cite the Same Websites for High-Stakes Consumer Decisions?
5. LLM Authority Index: How Stable Are AI Citations for High-Stakes Consumer Decisions?
6. LLM Authority Index: Reddit's Decline in AI Citations for Credit, Debt, Banking and Lending Decisions
7. Ahrefs: The 50 Most-Cited Websites in Copilot, September 2026
8. Tinuiti: AI Citation Trends Report Q3 2026
9. Microsoft Learn: Data, privacy, and security for web search in Microsoft Copilot
10. Machine Relations: How Concentrated Are AI Citations?

Related LLM Authority Index Research

Parent studies and measurement framework

• 2026 AI Citation Authority Study
• Credit, Debt, Banking and Lending Citation Authority Study
• Microsoft Copilot Platform Study
• Persistence-Portability Gap
• Do AI Platforms Cite the Same Websites?
• How Stable Are AI Citations?

Sibling Credit, Debt, Banking and Lending platform studies

• Google AI Overviews
• Google AI Mode
• ChatGPT
• Perplexity
• Gemini

Reddit trend studies for this category

• Credit, Debt, Banking and Lending Reddit parent study
• Google AI Overviews Reddit trend
• Google AI Mode Reddit trend
• ChatGPT Reddit trend
• Perplexity Reddit trend
• Gemini Reddit trend
• Microsoft Copilot Reddit trend