# Which Comparison Sites ChatGPT Cites for Finance Products

> In September's ChatGPT and Google AI Mode answers to everyday money questions, two editorial desks carried the product answers and small affiliate sites barely appeared. What a listing is worth, and what vetting each new site costs.

By Vlad Shvets, Partner @ Qvery · Published 2026-09-22 · Updated 2026-11-09 · 8 min read
Canonical: https://qvery.ai/blog/content-farms-chatgpt-finance-citations/

When someone offers to place a lender’s, a card program’s or a money app’s product on a long list of comparison sites so that AI assistants recommend it, someone on the team has to vet each site and check what it prints about your products.

**One quarter of comparison-site vetting, in our example**

| The quarter | By hand |
|---|---|
| New comparison sites to vet | About 25 |
| Time per site | About five hours: its pages, its disclosures, what it prints about you |
| Who does it | A partnerships manager and an analyst |
| Staff cost a quarter | $7,668 (about $30,700 a year) |

> 💳 At Qvery, we help lenders, card programs and money apps spend less on comparison-site listings that never reach the answers and keep their partnerships team on the sites that do, with custom AI agents that vet each new site, check what it prints about your products and draft the correction for your team to send.

Market research analysts and marketing specialists earned a mean $89,490 in May 2025, the Bureau of Labor Statistics reports; with benefits at 30% of private employer costs, that is about $127,800 a person.

(An interactive calculator on the page: roi-comparison-site-vetting.)

On those numbers the vetting costs $30,672 a year, and $13,735 with AI agents taking 65% of it, our assumption ($10,735 of staff time left and $3,000 a year to run them): $16,937 and 324 hours of staff time handed back, and every $10,000 of the build earned back in 7.1 months.

The listing fee is a cost of its own, and it buys nothing on a site the answers never cite.

## Where the Product Answers Come From

Our research, September 2026: ChatGPT and Google AI Mode answers to everyday consumer money questions, mostly set in the US. A cited page is one linked as a source under the answer; the shares are of answers that cited any source.

Start with the product questions: which budgeting app, which card for someone rebuilding credit, which platform for a retirement portfolio.

NerdWallet appeared in **32.02%** of the answers that cited any source, and CNBC in **19.38%**, nearly always through its Select best-of lists (the rest linked the CNBC home page). One or the other appeared in **41.15%**.

**Two Desks Carry the Product Answers; Small Affiliate Sites Barely Appear** (% of product answers citing the site)

NerdWallet appeared in 32.02% of product answers and CNBC Select in 19.38%, while three small affiliate sites together came to 2.95%.

| Label | Value |
|---|---:|
| NerdWallet | 32.02% |
| CNBC Select | 19.38% |
| Finder | 6.04% |
| MoneySavingExpert | 4.63% |
| Bankrate | 4.63% |
| WSJ Buy Side | 4.49% |
| Three small affiliate sites | 2.95% |
| Which? | 2.39% |

ChatGPT and Google AI Mode together, September 2026. Share of product answers that cited any source and cited the site; an answer can cite several. Source: Qvery research: ChatGPT and Google AI Mode answers to everyday consumer money questions, September 2026

Three small affiliate sites, which compare many providers on templated pages, earn a fee when a reader signs up and say on their own pages that partners pay them, appeared in **2.95%** of the product answers, all on budgeting-app questions. The big desks and the small sites alike are paid by the providers they list.

The leader changes with the product line: in budgeting apps and cards the desks lead, in investing platforms and payment apps the providers' own pages do. In a desk-led category the work is the facts the desks print about you; in a provider-led category it is your own product pages.

Those shares come from one month of answers; a daily record of your customers’ questions shows whether they hold in your category.

GEO Agent: https://qvery.ai/geo-agent/

## Where Questions About How Money Works Go

The second kind of question comes before a customer is ready to choose: what an origination fee is and whether it is negotiable, how pre-qualification differs from pre-approval, and what APR, the yearly cost of a loan, means.

The US consumer-finance regulator's pages appeared in 48.37% of these answers, Experian's explainers in 21.95% and investor.gov, the US securities regulator's site, in 14.63%. A watch list built from these answers alone would miss nearly nine in ten product answers.

The route worth testing is your own explainer for the questions your customers ask before they buy, written so it agrees with the regulator’s page the same answer is likely to read. Then watch whether the how-it-works answers start citing it.

![Two panels: product questions read comparison desks and providers’ product pages, owned by partnerships; how-it-works questions read the regulator and explainers, owned by help content.](https://qvery.ai/blog/content-farms-chatgpt-finance-citations/illustration-two-routes-two-owners.webp)

*The sources on each side are the ones the September answers leaned on most for that kind of question.*

## What the Agents Check for Each New Site

The agents we build for your partnerships team never contact a site or sign anything: the partnerships owner decides any contact, and compliance reads any listing or partner agreement before anyone signs. They open a site's pages only where its terms allow automated reading; a site whose terms forbid it goes on a person's list with the answers that cited it.

They read your recorded answers and two source lists, each source tagged as a desk, a provider, a regulator or a community and given an owner; a government or regulator site appeared in 82.93% of how-it-works answers against 20.22% of product answers, so the two lists differ.

When a comparison site on neither list appears under your answers or asks you to pay for a listing, the agents check whether it compares several providers, discloses provider payments and names credentialed authors, quoting the sentence behind each check. They add its share of your product answers, the product lines it appears in, and whether it lists you with correct figures.

Each new site becomes one row with a suggestion (ignore it, watch it, or send it to partnerships), and a drafted correction where the site prints your rate or fee wrong.

## Bring Last Quarter's Listing Invoices and Source List

Bring your current source list, last quarter's new sites and the listing invoices behind them. The agents vet them beside your partnerships team's own read, and every contact stays with the team; setup takes about two weeks once access is in place.

What the agents may read and do is set in a written agreement, your data is processed where that agreement says, and every action the agents take is logged.

Have Qvery Build Your Comparison-Site Vetting Agents: https://calendly.com/qvery-vlad/30min

## Frequently asked questions



### Does paying a comparison site get my product into AI answers?

A citation cannot establish whether payment caused an assistant to recommend your product. Ask for the share of your category's answers that cite a site before signing.

### Can an AI agent tell an affiliate site from an editorial one?

The agents we build quote the payment disclosure they find on each site, and a person confirms the classification before anyone acts on it.

### Who in a finance company should own this?

Give partnerships the product questions, passing corrections for your own pages to the site owner. Assign how-it-works questions to whoever writes your help and education pages, with compliance reviewing each change.

### Does a citation mean customers will choose us?

A citation shows which page an answer pointed to, not whether it recommended your product. Record whether the answer names your company separately from whether a customer later applies.

### What does Qvery do with the comparison-site list?

We use the GEO Agent’s records to identify where product facts need attention and do that work with your team. We also build custom agents for recurring site checks, train their reviewers during the two-week setup, and count hours saved and classification errors avoided each week.

### Which comparison sites does ChatGPT cite for finance products?

In September 2026 research covering ChatGPT and Google AI Mode together, NerdWallet appeared in 32.02% of product answers citing any source. CNBC appeared in 19.38%, nearly always through Select best-of lists. One or the other appeared in 41.15%.

### Do small affiliate comparison sites show up in AI finance answers?

Barely. Three small affiliate sites that compare many providers on templated pages and say partners pay them appeared in 2.95% of the product answers, all on budgeting-app questions.

### Which sources do AI answers use for questions about how money works?

The US consumer-finance regulator's pages appeared in 48.37% of how-it-works answers, Experian's explainers in 21.95%, and investor.gov in 14.63%. These September 2026 shares cover ChatGPT and Google AI Mode answers that cited any source. Investor.gov is the US securities regulator's site.

### How often do AI answers cite a government site on finance questions?

Government or regulator sites appeared in 82.93% of how-it-works answers versus 20.22% of product answers citing any source. These shares cover ChatGPT and Google AI Mode. A watch list built from how-it-works answers alone would miss nearly nine in ten product answers.

### Do providers' own pages lead in some finance categories?

Yes. In budgeting apps and cards the editorial desks lead, while in investing platforms and payment apps the providers' own pages do. The work differs: the facts the desks print about you, or your own pages.

### What does vetting comparison sites cost a finance company?

In this post's example, about 25 new comparison sites a quarter at about five hours each: $30,672 a year at about $127,800 a person with benefits. With agents taking 65%, it falls to $13,735, and 324 hours come back.

### Can vetting agents read every comparison site automatically?

Automated reading stops where a site's terms forbid it. That site goes on a person's list with the answers that cited it. The agents open site pages only where the terms allow automated reading.

### What do vetting agents check on a new comparison site?

Whether it compares several providers, discloses provider payments and names credentialed authors, quoting the sentences they find. Each new site becomes one row with a suggestion: ignore it, watch it, or send it to partnerships.

### What happens when a comparison site prints our rate wrong?

The agents draft a correction for that row. The partnerships owner decides whether to contact the site, and compliance reads any listing or partner agreement before anyone signs.

### How should a finance company answer how-it-works questions?

With its own explainer for the questions customers ask before they buy, written so it agrees with the regulator's page the same answer is likely to read. Then watch whether the how-it-works answers start citing it.

### How do you start with comparison-site vetting agents?

Bring your current source list, last quarter's new sites and the listing invoices behind them. The agents vet them beside your partnerships team's own read, and setup takes about two weeks once access is in place.

### Do vetting agents contact comparison sites?

No. They never contact a site or sign anything. The partnerships owner decides any contact, and every action the agents take is logged.

### How should a finance company organize its AI source lists?

Keep two source lists: one for product questions and one for how-it-works questions. Tag each source as a desk, a provider, a regulator or a community, and give it an owner. The agents read these lists alongside your recorded answers.