You can check what AI says about your firm in about an hour a month. Write five questions your ideal clients would ask. Put each to at least three assistants, in fresh chats with web search on. Record who gets named, which pages are cited, and what’s said about you. One answer tells you little: in our study of 600 AI answers, two assistants’ lists for the same city overlapped by 18%, and rewording the question changed most of the list. The method below is built around that.
Why one search tells you nothing
Most advisors who try this type their own firm name into ChatGPT once, see a reasonable summary, and stop. That tests whether the assistant can describe you when asked. It doesn’t test whether it names you when a client asks for help, which is the question that brings in clients.
Answers also move. In our study, asking the same assistant in the same city for a “fee-only financial advisor” and then for “the best fee-only financial planners for retirement planning” gave lists that overlapped by 28%. In 3 cities, ChatGPT and Perplexity named no firm in common at all. To learn anything, you need several questions, several assistants and a record you can compare month to month.
Step 1: Write five client questions
Use your actual best clients as the model. Think of the last few who signed and what they said in the first meeting. Write the question each of them might have typed before they found you:
- A plain local question. “I’m looking for a fee-only financial advisor in [city]. Who would you recommend?”
- A situation question. “I’m 58, my wife and I have about $1.5 million saved, and we want to retire at 62. Who should we talk to in [city]?”
- A specialty question. “I have RSUs vesting at a tech company and need a fee-only advisor who understands stock compensation. Who’s good in [city] or virtually?”
- A fee-model question. “I want a flat-fee advisor in [city], not one who charges a percentage of assets. Which firms should I look at?”
- A name question. “What do you know about [your firm] in [city]? What do they charge and who do they work with?”
The first four test whether you get named. The fifth tests whether what’s said about you is right. Write them the way a client talks, in full sentences, not as keywords. Keep the wording fixed from month to month so the results compare.
Step 2: Pick the assistants
Use at least three. Different assistants read different parts of the web, so one tells you about one slice of it.
- ChatGPT, with web search on. In our study it leaned on fee-only directories and the SEC’s adviser database.
- Perplexity. It leaned on consumer finance sites, rankings and review sites.
- Google, through AI Overviews or AI Mode, since many people still start in a search box.
- Claude, Gemini or Copilot, if your clients are likely to use them. Run one of them at least every quarter.
If you only have time for two, use ChatGPT and Perplexity. In our study they read very different sources.
Step 3: Ask the way a stranger would
Assistants personalize answers to the person asking. You want the answer a stranger gets, not the one that knows you’re an advisor in that city.
- Start a fresh chat for every question. Earlier questions in the same chat shape later answers.
- Turn off personalization. Use a temporary or signed-out chat where the assistant offers one, and make sure saved memories and custom instructions are off.
- Don’t say you’re an advisor. Ask as a client would.
- Name the city in the question. Don’t rely on the assistant guessing your location.
- Check that web search ran. The answer should show sources. If it doesn’t, ask again with search on.
- Ask one follow-up. Answer the assistant’s narrowing questions the way your ideal client would, or add a detail (“We have about $800,000 and want a one-time plan”). Who survives the follow-up matters more than who makes the first list.
Step 4: Record the same things every time
A spreadsheet is enough. One row per answer:
| Column | What to write |
|---|---|
| Date | When you asked |
| Assistant | ChatGPT, Perplexity, Google, and so on |
| Question | Its number from Step 1 |
| Named? | Yes or no, and your position in the list |
| Firms named | Every firm, in order |
| Your pages cited | Which of your own pages appeared as sources |
| Other sources cited | Directories, lists, review sites, press |
| What it said about you | Fees, specialties, location, minimum, credentials |
| Errors | Anything wrong or out of date |
| Where the error came from | The cited page that says it, if you can find it |
Save the full answer text or a screenshot too. When an answer changes next month, you’ll want to see exactly what it said before.
Step 5: Read the results
After one round you’ll have 15 to 20 answers. Look for patterns, not single results.
You’re never named and your site is never cited
The assistants either can’t read your site or find nothing on it that matches. Check that your site isn’t blocking AI assistants and that your fees and specialties are in plain text: Is your website blocking AI assistants? walks through it.
Your site is cited but you’re not named
The assistant read you and chose someone else. Open the pages of the firms it did name. Usually they state something yours doesn’t: a price, a niche, a minimum, the city in plain words. That’s the gap to close.
The same competitor keeps appearing
Look at which of its pages are cited. It may have a page for your city, a profile on every directory the assistant reads, or a press mention on your niche. Our guide on why AI seems to recommend big firms over boutiques explains why that’s coverage, not size.
The same directories keep appearing
Those are the sources that matter in your market. Make sure you’re on them, with the same fees, specialties and address as your site.
A fact about you is wrong
An old fee, a former partner, the wrong city, “fee-based” instead of fee-only. Find the cited page that says it and fix or remove it. Correcting the assistant in the chat changes nothing for the next person who asks.
You’re named on the name question, but nowhere else
The assistant knows you but doesn’t connect you to client situations. That’s a content gap, not a reputation gap: you need pages that answer the situation and specialty questions directly.
Step 6: Repeat monthly
Run the same five questions on the same assistants once a month, and add one new question each round from a recent prospect conversation. Answers vary from one run to the next, so judge by the trend over three months, not by one result. When you change something, such as publishing a fee page or fixing a directory profile, note the date in your sheet. That’s how you’ll see whether it worked.
Mistakes that skew the results
- Only asking your own name. It tells you about accuracy, not about being found.
- Asking in your everyday chat. Memory and history tilt the answer toward you.
- Changing the wording each month. Then you can’t tell whether the answer moved or your question did.
- Reacting to one answer. Expect some noise. Act on what repeats.
- Ignoring the sources. Who got named is the result. The cited pages are the cause, and the only part you can change.
- Forgetting the check clients run next. 40% of answers in our study told readers to look the advisor up on the SEC’s adviser database. Make sure your Form ADV says what your website says.
The done-for-you version
Wellspent members get this check every month in the AI visibility report. We write the questions a member’s ideal clients would ask, from what the member has published: specialties, location, fees and the clients they serve. We put them to an AI assistant with web search on, the way a client would, and pull out every advisor and firm named and every source cited. The report shows whether you were named, who was named instead, and which pages the answers leaned on.
Membership is $89 a month. It also includes a verified record in the public graph at find.wellspent.co, introductions from prospects with no lead fees, and quarterly press placement with media partners.