Which facts about a fee-only firm go with being named by ChatGPT and Perplexity? We took the 6,329 fee-only firms in the public graph with an advisor within 25 miles of one of the 85 US cities in our October AI study, and checked each against what the two assistants named there. 987 firms, or 16%, were named at least once. Membership in NAPFA or XYPN, a CFP on staff and a price published on the firm’s website all went with being named far more often. A niche specialty on the website, years in the business and SEC registration didn’t, once the other facts were accounted for.
These are associations in one day of answers, not proof of cause. A firm that publishes its price may differ from one that doesn’t in ways we can’t see.
How we ran it
On October 31, 2026 we asked ChatGPT, with web search, and Perplexity three questions in each of 85 US cities: a fee-only advisor, a fee-only planner for retirement, and a flat-fee advisor who doesn’t charge a percentage of assets. That gave 510 US answers, and we recorded every firm and advisor each one named.
For this study we started from the firms rather than the answers. Every fee-only firm in the public graph with an advisor within 25 miles of a study city is in the sample, whether or not an assistant named it. We marked a firm as named if any answer for a nearby city named the firm or one of its advisors, then compared naming rates across the facts we hold: what the firm’s website says, its Form ADV, its advisors’ credentials and the networks it belongs to. We report groups of at least 30 firms. The full method and its limits are at the end.
Source: Wellspent analysis of the October 31, 2026 AI study (510 US answers)
Networks went with being named most
NAPFA membership was the strongest fact in the data. 39% of NAPFA member firms were named at least once, against 11% of the rest. XYPN members were named 35% of the time. Firms in neither network were named 9% of the time, and they made up 77% of the sample.
That fits what the assistants read. In the first study, ChatGPT cited napfa.org in 61% of its answers and XYPN’s advisor search in 17%. The link held for both assistants and after accounting for city, size, credentials and website facts. Membership also comes with things that help on their own: a public profile with fees and specialties, and often a CFP.
A published price went with being named twice as often
We read the websites of 3,718 of the firms and kept only facts backed by a quote on the page. 29% of firms whose site published a price were named, against 15% of firms whose site didn’t.
The obvious objection is the flat-fee question, which asks for exactly the kind of firm that publishes a price. Leave it out and the gap holds: 23% against 12% on the fee-only and retirement questions alone.
Share named, with 95% interval. 3,718 firms whose websites we could read.
Source: Wellspent analysis of the October 31, 2026 AI study; website facts read October 28 to 30, 2026
Data
| What the firm's website says | Share named | 95% interval |
|---|---|---|
| Publishes a price | 29% | 27% to 31% |
| No price | 15% | 14% to 17% |
| Says it works virtually | 32% | 29% to 36% |
| Doesn't say | 18% | 17% to 20% |
| Says who it serves | 23% | 22% to 25% |
| Doesn't say | 17% | 15% to 19% |
| Names a niche specialty | 23% | 21% to 25% |
| Only retirement, tax or estate | 18% | 17% to 20% |
The price gap was widest where firms had the least else going for them. Among firms outside NAPFA and XYPN, 18% of those publishing a price were named against 10% of those that didn’t. Among network members, who were named often either way, it was 43% against 38%. For firms under $100 million in assets, it was 27% against 12%.
Firms that said they work with clients virtually were also named more often, 32% against 18%, and that held with the other facts accounted for.
Specialties didn’t make a measurable difference
We expected a named niche to matter, and at first glance it does: 23% of firms naming a specialty beyond retirement, tax or estate planning were named, against 18% of the rest. But firms that name a niche also tend to publish prices and belong to networks. With those facts in the same comparison, the specialty itself added nothing we could detect.
Saying who the firm serves (retirees, physicians, business owners) showed a weaker link: 23% against 17% raw, and smaller with the other facts in. It held more clearly on the fee-only and retirement questions.
Our three questions were general ones. A client who asks for a planner for physicians or for stock options is asking a different question, and specialties may matter far more there. This study can’t tell.
Size mattered to Perplexity, not to ChatGPT
Across both assistants, bigger firms were named somewhat more often: 23% of firms managing $1 billion or more, against 13% of firms under $100 million. That average hides two different patterns.
ChatGPT named firms of every size at about the same rate, and if anything slightly less often above $1 billion. Perplexity named firms of $1 billion or more about 4 times as often as firms under $100 million. With the other facts held equal, the $1 billion mark nearly doubled the odds with Perplexity and cut them by about 40% with ChatGPT.
The sources explain it. ChatGPT leaned on fee-only directories that list small planners. Perplexity leaned on SmartAsset, U.S. News and other ranked lists, which favor larger firms. Among the names Perplexity gave that we couldn’t match to the fee-only records, 118 (counted once per city) matched a registered adviser managing $1 billion or more. For ChatGPT it was 13.
Most named firms are still small, because most firms are. 47% of the firms named had under $100 million in assets, and 55% of ChatGPT’s. See Why AI seems to recommend big firms over boutiques for what that means for a small practice.
How a firm charges
Firms that don’t charge a percentage of assets at all, according to their Form ADV, were named 33% of the time against 14% for firms that do. Part of that is the flat-fee question. Leave it out and the gap is smaller but still clear, 20% against 11%.
On the flat-fee question itself, assistants mostly named firms whose sites said they charge a flat, hourly or project fee: 19% of those firms were named for it, against 7% of firms whose sites didn’t say so.
Experience, credentials and registration
- A CFP on staff went with being named: 21% against 6%. It held with the other facts accounted for, and among firms whose records come only from SEC data (13% against 6%), so it isn’t just a side effect of network membership.
- A CFA made no difference: 15% with one, 16% without.
- SEC or state registration looked like it mattered (17% against 14%) until size was accounted for. Then the difference disappeared.
- Firm headcount tracked assets: 23% of firms with 10 to 49 employees were named, against 14% of firms with one or two.
All the facts together
To separate facts that travel together, we put the main ones into a single model that compares firms in the same city. The chart shows how much higher the odds of being named were for firms with each fact, with the others held equal.
What held: network membership, a CFP, a published price, a larger team and saying the firm works virtually. What didn’t: a niche specialty, registration type and, across both assistants combined, assets above $1 billion. The facts that held have something in common. Each is stated plainly in a place the assistants read and can check against another source.
One more pattern runs through the data. The more facts a firm’s website stated in a way we could quote, the more often the firm was named: 11% of firms with 0 to 3 quotable facts, 17% with 4 to 7, 26% with 8 to 11 and 34% with 12 or more. Most of that is explained by the specific facts above.
Most named firms are in the public graph
Of the 2,650 firms and advisors named across the US answers, counting each once per city, 65% matched a fee-only firm in the public graph, and 51% matched one with an advisor within 25 miles of the city asked about. For ChatGPT it was 76%, and for Perplexity 61%, which named more large firms outside the fee-only records. 60% of the named firms we could match had their own website cited in an answer for their city.
What this means for a fee-only firm
The data can’t show that any single step gets a firm named. It does show what the named firms had in common.
- Be listed where the assistants look. NAPFA and XYPN members were named far more often, and ChatGPT read those sites in most of its answers. Wherever you’re listed, keep the facts identical.
- Publish your price in plain text. It went with being named at every firm size, and most of all for firms outside the networks.
- Say how and where you work. Firms whose sites said they work with clients virtually were named more often, with the other facts held equal.
- Don’t count on a niche alone. A specialty on the website didn’t move general questions. It’s what lets a firm answer the specific ones, which this study didn’t test.
- Check both assistants. They rewarded different things. Perplexity leaned toward large firms, ChatGPT didn’t. See How to check what AI says about your firm.
Methodology
Answers. The 510 US answers from our AI recommendations study: three questions in 85 US cities, asked once of ChatGPT (with web search) and Perplexity on October 31, 2026. Firms and advisors named in each answer were listed by Claude.
Firms. US firms in the public graph that pass Wellspent’s fee-only screen, with at least one advisor listed within 25 miles of a study city (straight-line distance between US Census place coordinates). That’s 6,329 firms, each counted once. The 15 Canadian cities are left out: we hold too few Canadian firms with comparable data.
Named. A firm counts as named if any answer for a study city within 25 miles named the firm or one of its advisors. We matched names exactly after lowercasing and removing punctuation and common words such as “financial”, “wealth” and “LLC”. Names that matched more than one nearby firm were dropped. Strict matching misses some variants, which lowers the naming rate for every group. As a check that doesn’t depend on names, we repeated the main comparisons using whether a firm’s own website was cited in a nearby answer. Networks, price, virtual service and the CFP showed the same pattern.
Facts. Website facts come from our reading of firms’ sites on October 28 to 30, 2026, keeping a fact only if its supporting quote appeared on the page; 3,718 firms had a site we could read, and website comparisons use only those firms. Assets, headcount, fee types and registration come from Form ADV. Credentials and years in the business come from the public graph’s advisor records (SEC data and network profiles). Network membership means an advisor at the firm appeared in NAPFA’s or XYPN’s public advisor search when we built the record.
Comparisons. We report the share of firms named in each group, for groups of at least 30 firms. The combined model is a logistic regression with each firm’s nearest study city’s naming rate as an offset, so it compares firms against their local baseline. Intervals are 95%.
Limits. This is one day of answers, and answers vary from run to run. It’s correlational: firms with these facts may differ in ways we don’t measure, such as reviews, press or how long their website has existed. The 25-mile rule includes suburbs, and in close pairs of cities, such as Dallas and Fort Worth, a firm named for either counts. Firms without a readable website are left out of the website comparisons. Wellspent membership isn’t analyzed: there are too few members to compare. Wellspent runs find.wellspent.co, one of the sites ChatGPT cited.