Finance AI Visibility Study 2026

01 / 07

Half of finance is in the dark.

People now ask AI assistants where to bank, how to pay and who to trust with their money. We built a city from 2,219 finance companies, one tower each.

We built a city from 2,219 finance companies, one tower each.

Every tower has 60 windows, one for each AI answer about its market. A window lights up when the answer names the company.

Named in the answerTaller: higher overall score

Lights out

02 / 07

1,077 towers never light up.

No AI assistant named these companies in any of their 60 answers. They are outlined in blue. No question named a company, so every mention was the AI’s own choice.

For about half of the companies in the study, a customer who asks AI for options will not see those companies listed.

Size, not sector

03 / 07

The big names shine.

Gold towers are the S&P 500 financials. One in five of them is never named, close to the 23% for the other companies in the US AI Visibility Study.

Blue towers are the Money20/20 attendees, most of them smaller firms. Half of them are never named. In finance, AI names the biggest companies far more often, whatever their sector.

Four assistants

04 / 07

Each side of a tower is one assistant.

SWIFT is the most named company in the study. Watch its tower turn: each side lights only where that assistant named it, one floor per question.

How often each assistant names the company a question is about:

OpenAI4.4% of answers
Claude5.3% of answers
Perplexity5.0% of answers
Gemini7.8% of answers

Rarely together

05 / 07

Often only one assistant names you.

Gold floors are questions where just one of the four named the company. That is 50% of the questions where any of them did.

On any single question, being named by one assistant says little about the other three.

Districts

06 / 07

Every part of finance has dark blocks.

The share of companies in each group that no AI answer named:

Banking53% dark
Cards and payments53% dark
Lending and credit41% dark
Investing and wealth50% dark
Insurance42% dark
Crypto and digital assets35% dark
Financial technology and data42% dark
Advisers and services57% dark

Find your tower

07 / 07

Is your company lit?

Search for any company in the study. The camera flies to its tower.

Find your tower

Browse all 2,219 companies

Chapters: what customers ask AI about money

Money20/20 USA · Las Vegas · October 18 to 21

Come see us at Booth 11041

In the Money Hall, between the Meridian Stage and the ArrA.I.val Stage, across the aisle from Mastercard.

Show me the wayOpen the booth map

QR code to aeo-audit.rezolve.com/us/finance/booth

Scan to take the map with you

The short version

Six things to know

1About half of the finance companies we studied are never named by AI assistants (1,077 of 2,219).
2Across all 15 questions, the assistants tend to agree on which companies are visible. On any single question, they often differ.
3Technical setup matters. Of the least prepared websites, 69% are never used as a source by AI, against 46% of the rest. Content optimized for the questions people ask AI builds on that groundwork.
4AI names providers far more often while people ask what is out there (16.9% of answers) than when they are ready to apply (2.6%).
5AI builds its answers mostly from other websites. A company's own website is 1.4% of the sources AI uses, and 7.7% in answers that name the company.
6At checkout, 86% of AI answers give rewards or cash back as the reason for their pick.

The ranking

The 50 highest overall scores

Companies are ranked by overall score, which covers more than how often they are named. Each company’s questions are about its own market, so the score rewards how strongly a company owns that market’s answers, not how famous it is. Rows marked rarely named came up fewer than three times, so read their scores with care. Pick a sector to see its top ten.

Browse all 2,219 companiesor search by name, website or stock ticker from the header

What counts as being named

We never name the company. Some answers do.

We asked 15 questions about each company’s market, worded the way a customer would ask them. None named the company. These are real answers from the study.

Asked of all four AI assistants

Low-fee alternatives to online payment platforms for small business payments

OpenAI•••

Claude•••

Perplexity•••

Gemini•••

OpenAI answered
“Here are practical, lower‑fee alternatives to mainstream online payment platforms (Stripe/PayPal) for small-business receipts, with short pros/cons and typical pricing so you can pick what fits your sales mix.…”
AI answer as returned, shortened and not checked for accuracy.

PayPalNamed by 3 of 4 AI assistants on this question

What else we found

Six more findings

1.4%

of the sources AI used are the company's own website

1.4% of the sources AI used were the website of the company the questions were about. 1,087 companies were never used as a source at all. The share is 7.7% in answers that name the company, and 0.8% in answers that do not.

What this means: AI mostly builds its answers from other websites. It uses a company's own website more when it is already naming that company. Companies whose content answers the questions people ask AI tend to be named more often.

Each square is a thousandth of the 669,453 sources. The 14 in blue are the companies' own websites.

28%

of websites have a guide written for AI (llms.txt)

85% of websites have a rules file (robots.txt) and 79% have a sitemap, the two files search engines need. The newer files written for AI tools are less common. 623 companies publish an llms.txt, a short guide that tells AI what is on the website.

What this means: More websites are set up for search engines than for AI tools. The files written for AI are the newer ones, and fewer companies have added them.

4 of 14

technical checks passed by the typical company's website

We ran 14 technical checks on each website to see how ready it is for AI tools. The typical website passes 4. The best in the study passes 13.

What this means: No company passes every check, so this is still open ground for everyone.

4 typicalbest 13 of 14

6.5x

more likely to be named when people ask what is out there than when people are ready to apply

When people ask what is out there, AI names a company in 16.9% of answers. When people are ready to apply, it is 2.6%.

What this means: AI names providers most often while people are still looking around, and less once they are about to decide.

14%

of the websites whose rules we could read name an AI crawler

A website can tell the automated readers AI companies send, called crawlers, what they may read. 259 of the 1,794 websites whose rules we could read name an AI crawler. The rest have only general rules, the kind written for search engines.

What this means: Most companies have not decided what AI may read on their website. Rules written for search engines are deciding for them.

259 of 1,794 readable robots.txt files name an AI crawler.

65%

of companies have linkedin.com among the sources AI used

linkedin.com is a source in answers about 1,437 of 2,219 companies, more than any other website. en.wikipedia.org is a source for 1,217.

What this means: What other websites say about a company reaches AI answers more often than what the company says about itself.

linkedin.com is a source for 1,437 of 2,219 companies.

By type of question

AI names companies when nobody is choosing yet

We asked seven types of question about every company’s market, shown here in the order a customer might ask them. AI names companies most when people ask what is out there, and least when people ask how to fix a problem.

16.9%

3.5%

4.6%

2.3%

5.5%

2.6%

1.2%

Browsing

5,744 questions

Specific need

4,437 questions

Alternatives

5,179 questions

Comparing

4,591 questions

Price

4,379 questions

Ready to apply

4,517 questions

Fixing a problem

4,434 questions

Exploring

Deciding

After buying

Browsing means asking what is out there, such as “Easiest way to buy bitcoin for individual investors”. Fixing a problem means asking how to solve something that went wrong. Comparing and ready to apply are the two types closest to a decision, and both are in the lower half.

By AI assistant

Gemini names companies most often. Perplexity lists the most sources.

How often each AI assistant names the company

Gemini

Flash with search

of 33,281 answers

Claude

Haiku with web search

of 33,281 answers

Perplexity

Sonar

of 33,281 answers

OpenAI

GPT mini with web search

of 33,281 answers

Who lists the most sources

Perplexity57.0%
Claude22.8%
Gemini10.6%
OpenAI9.6%

These shares come from a sample of the sources that we could match to an AI assistant. Perplexity lists far more sources than the others, so the most used websites in this study mostly reflect where it looks.

AI readiness

How ready these websites are for AI tools

AI tools are starting to read websites and act for customers, for example to compare accounts or start an application. We ran 14 technical checks on each website to see how easy that is. The typical website passes 4, and the best passes 13. The detail below is for your technical team.

Companies by checks passed

185

180

572

698

349

127

Checks passed, of 14

Each check, share of all 2,219 companies passing

Content Accessibility

Token budget (page weight)

97.3%

llms.txt published

28.1%

Markdown content negotiation

7.4%

Discoverability

robots.txt published

85.1%

XML sitemap

79.4%

HTTP Link headers (RFC 8288)

42.5%

Bot Access Control

Explicit AI bot rules

11.5%

API / Auth / MCP

OAuth Authorization Server discovery

5.2%

OAuth Protected Resource discovery

4.6%

API Catalog (.well-known/api-catalog)

1.1%

WebMCP tools

0.3%

MCP Server Card

0.2%

Agent Skills declared

0.1%

A2A Agent Card

0.0%

What AI is allowed to read

Most companies have not told AI what it may read

A website can tell the automated readers AI companies send, called crawlers, what they may read. We looked at the rules file (robots.txt) of 1,794 websites, and 259 name an AI crawler. The rest have only general rules, the kind written for web search. Companies that do set rules block the crawlers that collect pages to train AI more often than the ones that fetch a page to answer a question. The table below is for your technical team.

Training AI crawlers

Collect pages to train AI

websites block each one, on average

GPTBotnamed 219 · blocked 39
ClaudeBotnamed 200 · blocked 30
Google-Extendednamed 195 · blocked 29
CCBotnamed 161 · blocked 46
Applebot-Extendednamed 145 · blocked 25
Meta-ExternalAgentnamed 123 · blocked 31
anthropic-ainamed 108 · blocked 25
Bytespidernamed 108 · blocked 44
FacebookBotnamed 74 · blocked 20
Omgilibotnamed 53 · blocked 26
Diffbotnamed 39 · blocked 30

AI search crawlers

Index pages for AI search

websites block each one, on average

PerplexityBotnamed 173 · blocked 14
OAI-SearchBotnamed 134 · blocked 8
Amazonbotnamed 114 · blocked 25
YouBotnamed 73 · blocked 13
Claude-SearchBotnamed 68 · blocked 8
Applebotnamed 44 · blocked 4

Answering questions crawlers

Fetch a page when a person asks

websites block each one, on average

ChatGPT-Usernamed 139 · blocked 11
Claude-Usernamed 75 · blocked 10
Perplexity-Usernamed 74 · blocked 9
DuckAssistBotnamed 59 · blocked 9
MistralAI-Usernamed 22 · blocked 9

AI agents

What do these websites offer an AI agent?

Being named is the first step. The next is an AI agent sent to act for a customer. We scanned the websites of 2,200 of these companies for what an agent needs: a published way to pay or check out, and an interface built for agents.

2,195 of 2,200

394 companies

0.0%

of answers name the typical company

More than half of these companies were never named.

typical overall score

10overall score, line at the study’s typical 2860

Read the summary

AI assistants name the typical company in banking in 0% of answers, compared with 1.7% for the typical company in the whole study. Navy Federal Credit Union and Banco Popular de Puerto Rico are named most often in this sector.

See all 394 →

Cards and payments

375 companies

0.0%

of answers name the typical company

More than half of these companies were never named.

typical overall score

10overall score, line at the study’s typical 2860

Read the summary

AI assistants name the typical company in this sector in 0% of answers, compared with 1.7% for the typical company in the study. SWIFT, Coinstar, and Zelle are the companies in this sector that AI assistants name most often.

See all 375 →

Advisers and services

346 companies

0.0%

of answers name the typical company

More than half of these companies were never named.

typical overall score

10overall score, line at the study’s typical 2860

Read the summary

AI assistants name the typical company in this sector in 0% of answers, compared with 1.7% for the typical company in the whole study. The Wall Street Journal, The Fintech Marketing Hub, and Toptal are named most often by AI assistants.

See all 346 →

By category

Find your own kind of company

Each sector is made up of categories, 40 in all. Pick yours to see the companies in it and how they compare. Some categories hold only a few companies, so the count is shown beside each one.

Category	Companies	Answers naming the typical company	Never named	Typical overall score
Banking394 companies
Large national banks	61	1.7%	49.2%	27
Regional and community banks	148	0.0%	65.5%	27
Digital banks and neobanks	32	0.0%	59.4%	27.5
Credit unions	85	3.3%	41.2%	31
Small business banking and lending	38	1.7%	39.5%	29
Personal finance and money apps	30	4.2%	36.7%	28.5
Cards and payments375 companies
Card networks and card issuers	14	7.5%	35.7%	30
Payment processors, acquirers and point of sale	203	0.0%	55.7%	27
Digital wallets and person to person payments	29	3.3%	34.5%	29
Buy now, pay later	15	16.7%	20.0%	37
Cross-border payments and remittances	84	0.0%	61.9%	27
Business payments and spend management	30	1.7%	46.7%	28.5
Lending and credit93 companies
Mortgage and home lending	10	3.3%	40.0%	35.5
Personal, auto and student lending	62	1.7%	46.8%	28
Credit bureaus and credit scoring	13	16.7%	23.1%	37
Comparison sites and financial marketplaces	8	18.3%	25.0%	38.5
Investing and wealth238 companies
Online brokerages and trading apps	20	10.0%	25.0%	36
Asset managers	20	0.0%	60.0%	28
Wealth managers and financial advisers	22	3.3%	27.3%	35
Robo-advisers and investing apps	6	9.2%	50.0%	37.5
Private equity and alternative investments	74	0.0%	59.5%	27
Venture capital and growth investors	59	0.0%	59.3%	26
Retirement and workplace benefits	10	1.7%	40.0%	30.5
Investment banks and capital markets	27	5.0%	37.0%	29
Insurance97 companies
Auto and home insurance	22	1.7%	50.0%	29
Life insurance and annuities	17	0.0%	58.8%	27
Commercial insurance and brokers	29	1.7%	34.5%	27
Insurtech	18	1.7%	38.9%	28.5
Health insurers	11	5.0%	27.3%	37
Crypto and digital assets102 companies
Crypto exchanges and trading	22	5.0%	27.3%	35
Stablecoins and digital asset infrastructure	80	1.7%	37.5%	29
Financial technology and data574 companies
Banking and fintech infrastructure	315	1.7%	47.0%	28
Embedded finance and banking as a service	19	0.0%	52.6%	27
Fraud, identity and compliance	171	1.7%	40.4%	30
Market data, ratings and exchanges	37	15.0%	21.6%	36
Accounting, tax and payroll	32	15.0%	18.8%	38
Advisers and services346 companies
Consultancies and professional services	177	0.0%	65.5%	27
Law firms	32	0.8%	50.0%	27.5
Research, news and media	73	3.3%	34.2%	31
Marketing, communications and recruitment agencies	64	0.0%	64.1%	27

By list

S&P 500 companies are named most often

The typical company among S&P 500 companies is named in 5.0% of answers, against 1.7% across the study. Average AI readiness runs from 24 to 28 across the rows.

List	Companies	Answers naming the typical company	Never named	Average AI readiness, out of 100	Typical overall score
Money20/20 USA attendees	2,053	1.7%	49.7%	28	28
S&P 500 companies	88	5.0%	19.3%	24	31
All other companies	125	3.3%	36.0%	25	29
All companies	2,219	1.7%	48.5%	27	28

A company is counted on every list it is on. 47 companies are on both lists and are counted in both rows, so the rows add up to more than 2,219. The Money20/20 row is the companies in this study that are on the Money20/20 USA 2026 attending companies list, as it stood on October 1, 2026.

How to read this

What the study can and cannot tell you

Things to keep in mind

1,461 of 2,219 companies were named fewer than three times and are marked "rarely named". Read their scores as a sign that AI seldom brings them up, not as an exact comparison with their peers.
29 websites turned away our AI readiness check on every attempt and score 0 for it. That reflects their rules for automated visitors, not what the website contains.
The ranking is by overall score, which covers more than how often a company is named. So the order is not a list of the most named companies.

Read how we measured this

2,219 companies

Where does your company stand?

Every company in the study has its own report: how often each AI assistant names it, who gets named instead, and how ready its website is for AI tools.