What happens when someone asks an AI assistant for a fee-only advisor in your city

When someone asks ChatGPT “I’m looking for a fee-only financial advisor in Denver, CO. Who would you recommend?”, it doesn’t consult a list. It searches the web, reads a couple of dozen pages, and names the firms those pages describe. In the answer we follow here, from our study of 600 AI answers, ChatGPT cited 22 pages on 17 sites and named five small Denver firms. It quoted fees, showed ratings from local listings, and finished with three questions to narrow the list. Across the study, answers named 7.9 firms on average.

We asked on October 31, 2026, in a fresh chat with web search on. We don’t name the firms here; what matters is how the answer was built.

Step 1: The question becomes a search

The assistant starts by deciding what to look for. The cited pages show what it looked for here: a fee comparison site’s Denver page, NAPFA’s pages explaining what a fee-only fiduciary is, Fee-Only Network’s Denver page, a consumer finance site’s list of the best advisors in Denver, and the public graph’s Colorado page.

Its first line set the terms: it would start with “firms that explicitly operate on a fee-only basis, rather than firms that use the broader term ‘fee-based.’” The assistant knows the difference and checks for it. A firm whose pages blur the two gives it a reason to skip ahead.

Step 2: It reads more firms than it names

The 22 pages it cited fell into a few kinds:

• Directories and lists: the fee comparison site, three NAPFA pages, Fee-Only Network, the consumer finance list and the public graph.
• Firm websites: the homepages of the five firms it named, some read twice under slightly different addresses, and one firm’s fees page.
• Firms it read but didn’t name: four other Denver firms’ websites.
• Stray pages: a LinkedIn post by one advisor, an apartment building listing, and an old slideshow about another firm.

Two lessons sit in that list. First, being read is not being named. Four firms had their sites opened and still didn’t make the answer, because something on another firm’s page matched the question better. Second, the assistant reads whatever turns up, including pages you forgot about. That old slideshow is the kind of page that puts a former fee or a former partner into an answer.

This answer was typical. In the full study, ChatGPT cited about 15 sites per answer and Perplexity about 9, and 77% of all answers cited at least one fee-only firm’s own website.

Step 3: It names firms and says why

ChatGPT named five firms, all small independent practices. Each got a short block with the same kinds of facts:

• A one-line fit. “Best for advice-only planning”, “Best for comprehensive planning”, “Best for established wealth management.”
• How it charges. For one firm it quoted three published planning packages by price. For another it quoted a 1% investment management fee with lower tiers for larger portfolios. For a third it said “flat-fee financial planning” and stopped there, because that’s all the page said.
• Credentials and memberships. CFP, an MBA, a NAPFA registration.
• Where and how to reach them. Neighborhood, street address, phone number, and whether meetings are in person or virtual.

Those facts came from the pages it read. The firm with published packages got its prices repeated word for word. The firm that only said “flat fee” got no number at all. The assistant can only say what the pages say.

Step 4: It adds a map, with ratings

Next to the list, ChatGPT showed the five firms on a map, with star ratings and review counts drawn from local listing sources. One firm had 14 reviews; another had none. That’s part of why Yelp turned up so often in our study: ChatGPT cited it in 32% of its answers. A business listing with the right address, phone and website is part of the answer whether or not you look after it.

Step 5: It sets a price benchmark

Before telling the reader how to choose, the answer showed a small table of “Denver median” prices for hourly advice, a one-time plan, a monthly subscription and an annual planning fee. It took those from the fee comparison site it read first, and they differ from what we measured: our fee report puts the Colorado median for ongoing planning at $6,000 a year, against a US median of $4,500.

The point isn’t which benchmark is right. It’s that the assistant gives the client a price frame before the client ever talks to a firm. If your own price is on your site, the client can set it against that frame. If it isn’t, the frame is all they have. More on that in Publish your fees: why price pages get cited.

Step 6: It tells the reader how to check

The answer listed five questions to ask any advisor: strictly fee-only with no referral pay, fiduciary at all times, planning without handing over investments, total yearly cost, and CFP status. It pointed to NAPFA’s directory as a place to confirm fee-only status.

This happens a lot. In the full study, 40% of answers told the reader to look the advisor up on the SEC’s Investment Adviser Public Disclosure site, and 25% mentioned NAPFA. The client arrives having been told to check your registration. A Form ADV that says what your website says makes that check easy.

Step 7: It asks the client to narrow it down

The answer ended with three questions:

1. Approximately how much do you have to invest, in five brackets from under $250,000 to $3 million and up?
2. What kind of help do you want: comprehensive planning, investment management, retirement income, a one-time plan, or ongoing wealth management?
3. Which fee arrangement do you prefer: flat, hourly, a percentage of assets, or no preference as long as it’s fee-only?

Then: “With those answers, I can narrow this to the best three Denver advisors for your situation.”

This is where the real choice gets made. The second answer is built from the same pages, read more closely. A firm that says it works with clients under $500,000, offers one-time plans, and charges a flat fee in plain text can be matched to all three answers. A firm whose pages say “comprehensive wealth management” can’t be matched to any of them. The 5W and Haute Wealth report covered by InvestmentNews in May 2026 made the same point from the other side: it named the follow-up as the biggest gap between an AI engine and a competent human fiduciary.

The same question, asked elsewhere

We put the identical question to Perplexity for Denver. It named five firms too, and none of them were on ChatGPT’s list. It read different sources: review and consumer sites, XY Planning Network’s Denver page, three firms’ own sites, and a multi-office firm’s Denver location page. One of its picks was a firm whose site ChatGPT had opened and passed over. One firm it chose was described as having “no stated account minimum and a published declining fee schedule, making it worth considering if you want transparent pricing.”

That’s the study’s pattern. Across all cities, the firms the two assistants named for the same city overlapped by 18%, and a change of wording within one assistant left 28% overlap. A client’s result depends on which assistant they open and how they phrase it.

What this means for your firm

Each step points to something a firm controls:

• The search: be on the directories and city pages assistants read in your market, with the same facts everywhere.
• The reading: put your fee model, specialties and the clients you serve on your own pages in plain text, and find and fix the old pages that still mention you.
• The naming: give the assistant a one-line reason to pick you. “Advice-only planning for teachers in Denver” is a fit line. “Holistic wealth management” isn’t.
• The map: keep your business listing’s address, phone and website current.
• The price frame: publish your own price.
• The check: make your Form ADV match your website.
• The follow-up: state your minimum, your fee arrangements and who you work best with, so you match the second answer too.

To see how your own firm fares, run the method in How to check what AI says about your firm. For the steps to fix what you find, see How to get your firm recommended by AI assistants.

Where Wellspent fits

A Wellspent record holds the facts this answer was built from, your fee model and prices, specialties, service area, credentials and the clients you serve, verified against SEC data in the public graph that ChatGPT cited in this answer. Membership is $89 a month and includes the monthly AI visibility report, introductions from prospects with no lead fees, and quarterly press placement with media partners.