AI search is eating the referral game. Here's what advisors do in the next 12 months.

A financial advisor in Florida got a phone call not long ago from a prospect who said: "I was searching for a U.S. expat advisor, and your name came up on ChatGPT."

Key takeaways

• 8.7% of high-net-worth investors already use AI to find advisors - 25.3% among those with $5M+ - and one in four affluent households starts the search on an AI tool, not Google.
• Referrals are alive, but 96% of referred prospects research you online before contact, increasingly inside an AI conversation.
• The next 12 months: verifiable public facts, question-answering content, presence where AI engines read, and monthly visibility testing.

On this page

• The referral now has a referee
• What the referee looks for
• The 12-month playbook
• The window
• Where these numbers have limits
• Where Spaces fits
• The bottom line
• The bottom line
• Frequently asked questions

That story, reported by Barron's and retold in a recent Wealthtender guide, is easy to read as a curiosity. It isn't. No Google search. No directory browsing. No comparing bio pages across five websites. One conversation with an AI, one recommendation, one phone call.

I build technology for advisors, so I've been watching this shift closely. The numbers say it has already happened.

The referral now has a referee

A Ficomm Partners and Absolute Engagement survey, covered by American Banker in June, found that 8.7% of high-net-worth investors already use AI to find financial advisors. Among investors with $5 million or more, that figure is 25.3%. Among investors under 45, 15%. Wealthtender's own study of 500 affluent households goes further: one in four now starts their advisor search on ChatGPT, Gemini, or Perplexity, not Google.

But here's the number that should end the debate for every advisor who believes their practice runs on referrals: 96% of prospects who receive a personal referral research the advisor online before making contact.

Read that again. The referral is alive. What's changed is what happens next. Before a referred prospect calls you, they validate you, and a growing share of them do that validation inside an AI conversation. "My accountant recommended Sarah Chen. What are her credentials, and what do her clients say?" The answer that comes back determines whether your phone rings.

Referrals aren't dying. They're being pre-screened by machines. The referral game now has a referee, and it isn't human.

What the referee looks for

The mechanics of AI-driven discovery are not mysterious. The industry calls this answer engine optimization (AEO) or generative engine optimization (GEO), but strip away the acronyms and the research points to a short list of signals. Every one of them is learnable.

Independent reviews beat your own testimonials, by a wide margin. ChatGPT explicitly downweights testimonials published on an advisor's own website, because they're curated by the business. Reviews on independent third-party platforms carry dramatically more weight, the same way AI tools recommend physicians from Healthgrades and attorneys from Avvo. And here's the opening: only 9.3% of financial advisors currently use testimonials in their marketing at all. The field is nearly empty.

There's a story that makes this concrete. An independent advisory firm was competing for a client against a wirehouse advisor from a nationally recognized firm. The prospect read the independent firm's verified third-party reviews, searched for the wirehouse advisor's testimonials, found none, and chose the independent firm. Documentation beat brand.

Specific beats impressive. AI tools can only recommend you for a type of client if your content explicitly says you serve that type of client. "I work with a wide range of clients across all life stages" is invisible to a machine matching a query like "fiduciary advisor for tech employees in Austin with RSUs." "I specialize in retirement income planning for widows in Austin" is findable. The specificity that feels almost awkward in a bio is exactly what makes you discoverable.

Structure beats prose. Complete profiles with standardized fields, FAQ sections marked up with FAQ schema, consistent credentials across your website, directories, LinkedIn, and your regulatory records. AI tools cross-reference. Inconsistency reads as doubt, and doubt means you don't get cited.

Clicks are no longer the scoreboard. A prospect can hear your name in an AI answer, read your reviews, confirm your credentials, and call your office without ever visiting your website. Advisors who judge their marketing purely on site traffic will systematically undervalue the channel that's about to matter most.

Checking your own visibility? Request the free readiness check. A completed readiness review can show examples of how your firm appears in AI answers at the time of the check. Request a review: Can ChatGPT find your firm?

The 12-month playbook

None of this requires becoming a technologist. It requires treating AI visibility as infrastructure, the way you treated your website fifteen years ago. Here is the work, in the order that compounds.

Month 1: Run your baseline

You cannot fix what you haven't measured, and most advisors have never looked.

Set aside one hour. Open ChatGPT, Gemini, and Perplexity, and ask the questions your ideal clients actually ask:

• "Who are the best fiduciary advisors for [your niche] in [your city]?"
• "I'm a [your typical client profile]. What should I look for in a financial advisor?"
• "What do you know about [your firm name]?"
• "Is [your name] a good financial advisor? What do clients say about them?"

Screenshot everything. What comes back, accurate or not, is your current AI reputation. Pay attention to three things: whether you're named at all, whether the facts are right, and which sources the answer cites. Those cited sources are the referee's reading list. That's where you need to exist.

Then audit your third-party profiles against it. Every directory where you have a presence, fill every field: specializations, fee structure, service area, FAQs. Incomplete profiles give the machine nothing to work with.

Deliverable at the end of month 1: a one-page baseline document. What the machines say, what's wrong, what's missing, and the list of surfaces they read.

Month 2: Fix your own ground

Before you build anything new, make your own property say the right things plainly.

Rewrite your homepage and bio around three statements: whom you serve, what you do for them, and where you work. Not adjectives. Sentences a machine can quote. "Fee-only RIA in Sandy Springs, Georgia, serving retirees and business owners within 50 miles" does more for you than a paragraph about your passion for holistic wealth.

Rebuild your FAQ page for the discovery stage. Not "what are your fees?" but the questions people ask before they know anyone's name: "Can a financial advisor help Amazon employees in Seattle plan around RSU compensation?" Answer each in two or three plain sentences, and mark the page up with FAQ schema so machines can parse the question-answer pairs directly.

Then reconcile every surface: your website, LinkedIn, your regulatory records, every directory. Same niche, same language, same credentials everywhere. AI tools cross-reference, and every contradiction is a reason not to cite you.

Deliverable at the end of month 2: a website and profile set that states your specialization identically everywhere, with machine-readable FAQs.

Months 3 and 4: Start the review engine

This is the highest-leverage move on the list, and the one almost nobody has made. Fewer than one in ten advisors collect client reviews at all.

The SEC Marketing Rule allows testimonials and reviews with proper disclosures, and those disclosure requirements turn out to align with how AI tools judge review credibility. A compliant process is also the winning process. Pick one independent third-party platform, build a simple ask into your client calendar (after annual reviews, after a plan milestone, after a genuine thank-you), and collect steadily rather than in one burst. Machines, like prospects, trust a stream more than a spike.

Two rules keep this clean: never write or edit a client's words, and never cherry-pick only the happy clients into the ask. Steady, honest, disclosed.

Deliverable at the end of month 4: a working review cadence and your first real body of independent, verified reviews.

Months 5 and 6: Publish answers, not brochures

AI tools cite sources that answer real questions well. Thin content gets filtered; useful, specific, honest answers get quoted.

Take the questions from your month-1 baseline, the ones your ideal clients ask before they know your name, and answer them properly. One question per piece. "How should a widow in Texas think about Social Security timing after 60?" is a publishable answer. "Our comprehensive approach to wealth" is not.

The bar is simple: could a machine quote your answer, verbatim, and look smart? Plain language, specific numbers where they help, no hedging filler. Two strong answers a month beats eight thin ones.

Deliverable at the end of month 6: four to six substantive, quotable answers live on your site, each matching a real discovery question.

Months 7 through 9: Get cited where the machines read

Your own site is necessary but not sufficient. AI answers lean on third-party sources, so spend this phase showing up on them.

Contribute expertise where advisors get quoted: industry publications, local business press, podcast interviews, professional directories with real depth. When a journalist or host needs a quote about retirement income or RSU planning, be the person who answers fast and plainly. Every one of those appearances is a citation the referee can find.

This is also the phase to revisit your baseline. Re-run the same queries from month 1 and compare. You are looking for movement: your profiles appearing as cited sources, your FAQ answers being paraphrased, your name surfacing in longer lists.

Deliverable at the end of month 9: a handful of third-party citations and a documented before-and-after on your baseline queries.

Months 10 through 12: Measure what matters and compound

By now the flywheel has inputs. Give it instrumentation.

Add one question to your intake process: "How did you hear about us?" and train everyone who asks it to probe when the answer is "online." Online means Google, or it means ChatGPT, and those are different channels with different playbooks. Advisors who ask carefully are already hearing "an AI recommended you" out loud.

Re-run your baseline quarterly. Double down on whatever surfaces moved. Fix whatever still contradicts. And keep the review engine and the publishing cadence running, because consistency is itself a signal.

Deliverable at the end of month 12: a repeatable system, an intake process that measures the channel, and a documented year-over-year change in what the machines say about you.

The window

Two years ago, the researchers behind that Ficomm survey didn't think AI was common enough to even include as a question. Today it has its own section. That's how fast this is moving.

The advisors who build this presence now, while fewer than one in ten of their peers have even started collecting reviews, are compounding an advantage that gets more expensive to close every quarter. The advisors who wait for certainty will get it. It will arrive in the form of a prospect who chose someone the machine recommended.

Where these numbers have limits

Survey percentages describe early adopters in specific samples, not your client base: an advisor serving retirees will feel this shift later than one serving tech executives. The direction is clear and the pace is local - test what AI says about your firm before rearranging your budget.

Where Spaces fits

Spaces is developing a way for advisers to work on AI-search visibility and LinkedIn outreach together. We cannot make an AI service recommend a firm or claim a live stream of exclusive, qualified prospects. The path from visibility to real introductions is being tested. Request the free readiness check for examples from a completed review: Can ChatGPT find your firm?

The bottom line

The referral now has a referee, and the referee is an AI. The advisors who win the next 12 months will be the ones an AI can verify and recommend. Start measuring what it says about you this month.

Frequently asked questions

1. Are people really finding financial advisors through ChatGPT?

Yes. A Ficomm Partners and Absolute Engagement survey found 8.7% of high-net-worth investors already use AI to find advisors - 25.3% among those with $5 million or more - and Wealthtender found one in four affluent households starts the search on an AI tool, not Google.

2. Are referrals dead for advisors?

No, but 96% of referred prospects research the advisor online before making contact, and a growing share do that validation inside an AI conversation. The referral is alive; what changed is the referee.

3. What should advisors do in the next 12 months?

Make your firm easier for people and AI tools to verify: keep public facts consistent, answer real client questions, and test how AI tools describe you. These steps do not ensure a recommendation.

Spaces is working on this problem for advisers: making firm information clearer and testing whether visibility and outreach lead to suitable conversations. If you want to know what AI tools currently say about your firm, request our free AI-search readiness check . The form requests a readiness review; a completed review may show examples from the time of the check, not a live score or promised introduction.