Updated on 19 October 2026, by João Leite, Founder of murmur.

For a fintech or a bank to appear in ChatGPT, the site needs to answer the customer’s questions with fee, term and authorization in writing. murmur, a GEO agency in Brazil that places brands in the answers of ChatGPT, Claude, Google AIO and other AI agents, has published sector studies on finance. The seven in this guide are from September 2026. In each one, murmur asked 25 questions from people looking for an account, a card machine, a consortium or a remittance, without the name of any company. In the consortium study, those who asked how the product works received an explanation and the advice to check the institution with the Central Bank. In these seven sectors, the AI cited the company whose site it had consulted far more often.

  • In the consortium study, those who wanted to understand the product received an explanation, not a recommendation. In murmur’s study, the 9 answers of this type hardly mentioned any company name and told people to check the authorization with the Central Bank.
  • The Central Bank was one of the most cited names in the consortium study. It appeared in 48 of the 75 answers. Its ranking of administrators was the most consulted site, in 21 of the 40 answers with the sites visible.
  • In the seven finance sectors in this guide, the consulted site turned into a name. With the site opened by the AI, the company name appeared 86% to 100% of the time; without it, 1% to 28%.
  • In the consortium study, the bank lost ground when the question narrowed. Some bank appeared in all 15 answers to the general recommendation and in 57% of the 21 answers from those asking for an independent administrator.

GEO for fintechs and banks: what changes when the subject is money

To appear in ChatGPT in finance, what changes is the subject: money is a topic where people check before deciding. Someone opening an account, asking for a loan or choosing a card machine wants to know the fee, the term and whether the Central Bank authorizes the institution. The AI that answers this person repeats what it finds written, and that is why the page needs to state these facts in full. The general step-by-step guide, for any company that sells to other companies, is in the guide on GEO for B2B companies, and the insurance one is in GEO for insurers and brokers.

How AI recommends banks and fintechs when the person is still getting to know the product

In murmur’s consortium study, those who asked how the product works received an explanation and the advice to check the authorization with the Central Bank. There were 9 answers to three questions: how much the administration fee costs, whether a consortium or a loan is better, and how to choose a safe administrator. In them, the AI explained the product and hardly recommended any company.

In the banking correspondent study, the question from those still deciding also ended up without a name. The 15 answers from those asking how to become a correspondent, how much you earn and whether you need certification did not mention any system or promoter among those cited in the study. In 26 of the study’s 74 answers, the AI cited no company at all. In 34, it cited two or more.

For the fintech, the question from someone still getting to know the product calls for another kind of page. It explains how the product works, how much it costs, what the risks are and where to check the institution. In these studies, it was this explanation that the AI wrote in the answer, and it needs pages that provide it ready-made. The split between digital account and traditional bank, in the business account study, is in the answer about how AI answers about banks and fintechs.

Central Bank, Reclame Aqui and industry association: who the AI cites besides companies

In murmur’s finance studies, the AI cited names that serve as references: the Central Bank, Reclame Aqui and the industry association. In the consortium study, the Central Bank appeared in 48 of the 75 answers and Reclame Aqui in 14. ABAC, the association of consortium administrators, appeared in 9. None of the three sells consortium, and all three serve as references for the AI.

The Central Bank was also the most consulted site in the consortium study. Its ranking of administrators was a source in 21 of the 40 answers with the sites visible, counting the three Central Bank URLs. The most consulted administrator site was a source in 5. In the banking correspondent study, the Central Bank appeared in 22 of the 74 answers, and Febraban, the federation of banks, in 12.

In the card machine study, Reclame Aqui and the equipment manufacturer entered the text of the answers without competing for the sale. Reclame Aqui, which is not written by the company, appeared in the answer alongside the companies. How the AI uses the Reclame Aqui score and solution index is in the guide on reviews and Reclame Aqui in AI.

Bank or specialized company: what changes when the question narrows

In murmur’s consortium study, bank and independent administrator appeared together in the open question, and the bank lost ground when the question narrowed. Some bank appeared in all 15 answers from those who just wanted the best administrator, and so did some independent administrator. In the 21 answers from those who asked for an independent administrator, digital or not tied to a bank, the bank dropped to 57%.

For a fintech, the space may be in the question that asks for what it has that is different: outside the bank, digital or for a customer profile. Among the 25 in the consortium study are the one from those who want to sign up 100% online and the one from the MEI and the self-employed worker without a payslip. Each one calls for a page of its own, with the words of the person asking and the answer in the first paragraph. A new page is justified when the customer’s question actually changes, and it brings the complete answer. Written in other words, the same question stays on the page that already exists. What the narrower question brought in the other sectors murmur measured is in the guide on GEO for startups.

The consulted site turned into a name in the answer in seven finance sectors

In seven finance sectors that murmur measured, the company appeared far more in the answer from ChatGPT, Claude and other AI models when its site had been consulted. The count uses only the answers that show the consulted sites: those from ChatGPT and Claude in all sectors, plus those from Perplexity in card machines, international remittance and recurring billing, and those from Google AIO only in billing. In business accounts, it was 93% with the site open and 15% without it, in 47 of 75 answers with the sites visible. In card machines, 98% and 28% (57 of 100); in international remittance, 86% and 10% (30 of 100); in recurring billing, 94% and 16% (63 of 99). In consortium, 100% and 16% (40 of 75); in identity verification and anti-fraud, 95% and 3% (29 of 75); in banking correspondent, 97% and 1% (36 of 74). The numbers show reading and citation together, not that one causes the other; why the site matters, alongside reviews and the press, is in where AI checks the B2B supplier.

What a financial product page needs to include to appear in ChatGPT

The product page that wants to appear in ChatGPT includes in text the fee, the term, what is included and who authorizes the institution. A fee with the value, a term in days and the name of the authorized institution are facts the AI can copy into the answer. The fee goes in the page text, not only in an image or a fees PDF. In Vercel’s measurement, a company that hosts sites, from December 2024, OpenAI’s crawlers and Anthropic’s ClaudeBot downloaded the JavaScript and did not execute it, and what a simulator calculates on screen with JavaScript stayed outside what they read. The measurement has not been repeated, and today’s test is to open the page source code (Ctrl+U in the browser) and look for the fee, as the guide on text ready in the HTML shows.

An old fee on the site can become an old fee in the answer. That is why the product page brings the date of the last fee change next to the value. The page the AI cites states facts: fee, term, product rule and authorization. A promise of return or approval does not help the AI answer.

Old name, group and brand: the AI still writes the name the company left behind

In murmur’s international remittance study, the AI still wrote the old name of a company that had been renamed, in 4 of the 100 answers. In the same study, other names on the list belonged to the same house, through acquisition or a name change. In the identity verification and anti-fraud study, one of the companies cited had changed its name, and two of the most cited belong to the same group.

A fintech or bank that has changed its name or owner writes on its site the current name, the old name and its group. The same care applies to official registrations, social media profiles and third-party pages that cite the company. When the AI mixes up two institutions, the steps to fix it are in the guide on correcting what the AI writes wrong about the company.

How to measure the presence of a fintech or a bank in AI models

To measure a fintech’s presence in AI models, repeat the customer’s questions in rounds and count the answers that mention its name. Presence is how many answers the brand appears in, divided by the total number of answers. The same question goes to ChatGPT, Claude, Perplexity and Google AIO, because each AI model has its own list. The answer changes from one round to the next, and no institution has a fixed place in it.

In finance, the calculation separates three types of answer. One recommends the company, another only explains the product, and the third tells the person to check with the Banco Central. The question from someone who wants to understand the product, which almost never mentions a name, is counted separately, so it does not lower presence on recommendation questions. Also note the sites the AI consulted and whether the company’s own site was among them.

How many questions to use is in the guide on how many questions to measure brand presence. The spreadsheet that adds up presence, answer by answer, is in the step-by-step guide to measuring presence in ChatGPT. Timeline and price have their own guides: how long it takes to appear in ChatGPT and how much a GEO agency costs.

murmur helps correct what the AI says wrong and tracks the answer every week

murmur helps the company correct what the AI says wrong, such as an outdated price, an old name or a company confused with another. The work is to correct and position the brand the right way in AI models, not just to show the number. A specialist follows the company from the diagnostic onward and studies the questions its customer asks. From those, murmur writes the guides that answer each one. murmur’s case studies show this work with each client’s name and the date of each measurement.

murmur publishes the guides on a subdomain with the company’s brand, and all the content stays with the company when the contract ends. Week after week, murmur asks the same questions to the AI models again, and the change in the answer after a correction shows up in the week it happens. The questions are asked by a dedicated murmur server, installed in Brazil, with the browser in Portuguese and a single internet address, the same every week, and never a team member’s computer. That way, the city where the measurement is made does not affect what the AI answers, and the week of the correction can be compared with the previous one. The company tracks this change, and that of its competitors, in a dashboard that comes with the service. The work takes little time from the company’s team, and the contract has no minimum term, no exit penalty and no setup fee.

Frequently asked questions

Does the AI recommend a bank and a specialized company the same way?

Not always: in murmur’s consortium study, the bank shared the open question with the independent administrator and lost ground when it was asked for. A bank appeared in the 15 answers to the general recommendation and in 57% of the 21 that asked for an independent administrator. The split between digital account and traditional bank is in the guide on how many Brazilian companies appear in AI answers.

Can the AI recommend an institution that is no longer authorized?

It can: in murmur’s international remittance study, the AI recommended a currency broker that the Banco Central liquidated in April 2026. It was in 1 of the 100 answers, for someone who wanted to send a large amount. The study does not show where the AI got the information. The steps to request a correction are in the guide on correcting what ChatGPT says wrong about the company.

Does Reclame Aqui influence what the AI says about my fintech?

In murmur’s finance studies, Reclame Aqui appeared in the answers as a reputation reference, alongside the companies. In the consortium study, it appeared in 14 of the 75 answers, and in the card machine study it entered the text without competing for the sale. What the AI repeats from the score and the resolution index is in the guide on reviews and Reclame Aqui in AI.

Do I need a comparison page with the banks?

It depends on the sector: in murmur’s card machine and remittance studies, 6 of the ten most consulted sites were third-party sites, such as comparison sites and blogs. In the recurring billing study, 9 of the ten were the platforms’ own sites. What a comparison page needs to have is in the guide on comparison page and alternatives page.

In a regulated sector, GEO is done with a fact page, and that is where the AI takes the answer from. Fees, timelines, product rules and authorization are the sentences the AI repeats and the ones legal can check. Promises of returns or approval stay out, and the AI does not need them to answer. In insurance, where the regulator is SUSEP and the question is about coverage and claims, the steps are in how to do GEO for insurers and brokers.

Where to start

Start by asking ChatGPT, Claude, Perplexity and Google AIO your customer’s questions, from the open-ended recommendation to questions about the product. Note who appears, which sites the AI consulted and whether it told the person to check with the Banco Central. murmur’s free diagnostic runs this first round from the fintech’s website URL, with no sign-up, in less than two minutes. murmur has published studies of finance sectors, each with the 25 questions and the calculation for each AI model. The seven in this guide are from the September 2026 round: consortium, card machines, business account, international remittance, recurring billing, identity verification and anti-fraud and banking correspondent.

Read also

References

  • Vercel, “The rise of the AI crawler”, 17 December 2024, with data from MERJ. Measurement on the Next.js site and on the Vercel network, never repeated. https://vercel.com/blog/the-rise-of-the-ai-crawler (accessed on 23 September 2026).
  • murmur, “O que as IAs respondem quando alguém procura uma administradora de consórcio”, sector study from September 2026: 25 questions, once in ChatGPT, Claude and Google AIO, 75 answers; 40 with the sites visible. https://murmur.marketing/estudos/consorcio-2026-09/ (accessed on 12 October 2026).
  • murmur, “O que as IAs respondem quando alguém procura plataforma para correspondente bancário”, sector study from September 2026: 25 questions, once in ChatGPT, Claude and Google AIO, 74 answers, of which 24 from Google AIO. https://murmur.marketing/estudos/correspondente-bancario-2026-09/ (accessed on 12 October 2026).
  • murmur, “O que as IAs respondem quando um lojista procura maquininha”, sector study from September 2026: 25 questions in 4 AI models, 100 answers; 57 with the sites visible, in ChatGPT, Claude and Perplexity. https://murmur.marketing/estudos/maquininhas-2026-09/ (accessed on 12 October 2026).
  • murmur, “O que as IAs respondem quando alguém procura mandar dinheiro para fora ou abrir conta em dólar”, sector study from September 2026: 25 questions in 4 AI models, 100 answers; 30 with the sites visible, 14 from Claude, 13 from Perplexity and 3 from ChatGPT. https://murmur.marketing/estudos/remessa-internacional-2026-09/ (accessed on 12 October 2026).
  • murmur, “O que as IAs respondem quando uma empresa procura como cobrar assinatura”, sector study from September 2026: 25 questions in 4 AI models, 99 answers, of which 24 from Perplexity; 63 with the sites visible. https://murmur.marketing/estudos/cobranca-recorrente-2026-09/ (accessed on 12 October 2026).
  • murmur, “O que as IAs respondem quando uma empresa procura onde abrir a conta PJ”, sector study from September 2026: 25 questions, once in ChatGPT, Claude and Google AIO, 75 answers; 47 with the sites visible. https://murmur.marketing/estudos/conta-pj-2026-09/ (accessed on 12 October 2026).
  • murmur, “O que as IAs respondem quando alguém procura verificação de identidade e antifraude”, sector study from September 2026: 25 questions, once in ChatGPT, Claude and Google AIO, 75 answers; 29 with the sites visible. https://murmur.marketing/estudos/kyc-antifraude-2026-09/ (accessed on 12 October 2026).
  • murmur, sector studies: https://murmur.marketing/estudos/; free diagnostic: https://murmur.marketing/en/diagnostic/ (accessed on 12 October 2026).