The first thing every $5M prospect does before they call you is open ChatGPT.
Not Google. Not a referral. Not the form on your homepage. They type the prompt — “best fee-only fiduciary advisor in Austin for someone with $3M in retirement assets” — and they read three paragraphs, click two of the cited links, and decide who they want to interview. The shortlist is set before your phone rings.
If your firm isn’t in those three paragraphs, the meeting never happens. There is no second-place click-through to recover from. The AI engine made the shortlist, and the shortlist was final.
This is not a forecast. This is the documented 2026 buyer behavior for the high-net-worth segment that drives RIA economics.
Why RIAs lose AI citation share by default
Three forces compound. First, advisor-niche templates (FMG Suite, Twenty Over Ten, AdvisorWebsites) inject schema and compliance disclosures client-side, where AI crawlers cannot see them. Second, the citation surfaces AI engines actually trust — NAPFA, XYPN, BrokerCheck, LinkedIn — are external directories most firms have never optimized. Third, FINRA Rule 2210 makes most marketing playbooks unusable, so the category has stayed empty.
The structural problem is that AI engines do not behave like Google. ClaudeBot, PerplexityBot, and the OpenAI training crawler GPTBot do not execute JavaScript. They fetch raw HTML and move on. Vercel logged 569 million GPTBot requests in a single month across its network in 2026, with AI bots accounting for 4.2% of all HTML page requests and GPTBot traffic up 305% year over year — but every one of those requests reads only the initial HTML response.
That detail is load-bearing for advisors specifically. Compliance-archive scripts, FMG Suite injectors, and most advisor-template schema plugins write JSON-LD to the page after the script tag executes in the browser. Googlebot renders the page and sees it. ClaudeBot does not. The result is a site that looks fine in Search Console and is structurally invisible to the engines deciding who shows up when a $5M prospect asks for recommendations.
The 22 percentage-point citation gap between fully-populated schema and sparse schema (Growth Marshal, February 2026) doesn’t apply if your schema isn’t in the initial HTML at all. You’re not in the comparison.
What changed for RIAs in 2026
The numbers came in two waves and they both pointed the same direction.
The buyer wave came first. Wealthtender’s 2025 study of 500 high-income adults — published widely through early 2026 — found that 1 in 4 plan to use ChatGPT or Gemini to find their next financial advisor. Menlo Ventures’ 2025 State of Consumer AI report (cited by Edge Partners through April 2026) put the share of $100K+ households using AI tools regularly at 74%. Households earning $100K+ are the funnel into the $1M+ investable segment that drives RIA AUM growth. The buyer is already there. The buyer is HNW. The buyer is asking AI directly.
The advisor wave came second, and it confirmed the structural inevitability of the shift. Schwab’s RIA AI Adoption Study, published January 22, 2026 with a sample of 533 RIAs, found that 63% of RIAs themselves now use AI in some capacity. Within that group, 59% believe AI will directly impact client relationships within one year and 68% expect AI to be transformative to financial advice within three years. The supply side is moving as fast as the demand side. The category will get crowded.
The arithmetic on a single citation win is what makes the urgency unambiguous. A $5M AUM client at the typical 1% AUM fee returns $50,000 per year, recurring. The site rebuild that gets you cited is a one-time cost. One client pays it back fifty times over. There is no other channel in the advisor stack — not paid search, not LinkedIn ads, not COI dinners — where the LTV-to-CAC math is even close to that.
The constraint is that you have to be in the citation set before any of that compounds. Right now, most RIAs aren’t, because nobody has measured what the citation set actually looks like.
The 5 directories ChatGPT actually cites for advisors
There is no published per-vertical citation share study for financial advisors as of May 2026. The 5W AI Platform Citation Source Index 2026 — the most authoritative cross-engine citation map in the field — is cross-vertical only. It tells you Wikipedia accounts for roughly 5% of all individual ChatGPT citations and shows up in 18% of cited conversations, and that Reddit accounts for roughly 3% of citations and 13% of conversations. It does not break out which directories dominate when the prompt is “find me a fiduciary in Phoenix.”
What the 2026 advisor-marketing literature does establish, with named sourcing, is the directory short list AI engines pull from when an HNW user asks the question.
NAPFA — Find an Advisor. The National Association of Personal Financial Advisors’ fee-only finder is explicitly named in Edge Partners’ April 2026 analysis of the $5M-prospect scenario as a directory ChatGPT routinely surfaces. NAPFA membership requires a fiduciary commitment — exactly the filter HNW prospects are trying to apply. AI engines have learned this and route to NAPFA when prompts include “fee-only” or “fiduciary.”
XY Planning Network (XYPN). Also named in the Edge Partners 2026 review. XYPN’s adviser search is the next-gen-advisor counterpart to NAPFA — it skews younger, monthly-retainer, and specialty-focused. AI engines surface XYPN heavily on prompts that include lifecycle specifics (RSU planning, equity compensation, early-career physician).
LinkedIn headlines. Wealthtender’s 2026 advisor-AI guide confirmed LinkedIn as a citation surface and was specific about why: a profile headline reading “Fee-only CFP in Boston specializing in physicians” measurably outperforms generic “Wealth Manager” headlines in AI retrieval. The headline is the entity-graph signal LLMs latch onto, because it concentrates the specialty + credential + geography in one parsable string.
Google Business Profile reviews. Confirmed across the Wealthtender and Edge Partners reviews as a citation source AI engines pull from for local-services queries, including advisor queries. The detail that matters: AI engines weigh review content (specifics about a niche or outcome) above review count. A firm with 18 reviews mentioning “RSU planning” and “Bay Area tech founders” outperforms a firm with 80 generic five-star reviews.
BrokerCheck (FINRA). BrokerCheck is the regulatory disclosure surface for every registered representative. It is a Plan-listed citation hypothesis we are still empirically testing — the published 2026 reviews mention it as part of the advisor citation stack but do not yet quantify its share. The reason it deserves your attention now is that BrokerCheck is the AI engines’ fact-check surface. Even when it isn’t the cited URL, the engines triangulate against it before recommending you.
The four candidates the Plan-doc named that we have not yet seen quantified for advisor-specific AI citation share are CFP.net, SmartAsset SmartAdvisor, Zoe Financial, and Wealthmanagement.com. The honest position is that they are presumed citation surfaces awaiting measurement. The arithmetic of an RIA paying for SmartAsset SmartAdvisor leads is that the LTV per converted lead has always been bottle-line — a single $5M client justifies a year of the spend — but no public study has yet established whether the SmartAsset directory pages themselves get cited at the rate the buying funnel implies.
That gap is exactly the empty category we flagged in the research pack. The Wave 1 Wednesday data drop on this hub will be the first per-state advisor citation share scan: NAPFA vs. XYPN vs. SmartAsset vs. named-firm pages, measured across the top 25 metros. The data does not exist publicly yet. We are running it.
The 15 prompts $5M AUM prospects ask in 2026
The reason citation strategy starts with prompts — not with on-page SEO — is that AI engines retrieve against the intent of the prompt, not against the keyword. “Best fee-only CFP for physicians in Boston” and “find a fiduciary advisor in Boston who works with doctors” surface different shortlists, even though the keyword overlap looks identical. The prompt is the unit of optimization.
These are the prompts the 2026 advisor-marketing literature documents, sourced primarily from Edge Partners’ April 2026 review of the $5M scenario, Wealthtender’s 2026 buyer-side guide, and Reddit advisor-marketing analyses:
- “Who is a good fiduciary advisor in Austin for someone with $3M in retirement assets?”
- “Recommend a wealth manager in Chicago who specializes in equity compensation.”
- “Best fee-only CFP for physicians in Boston.”
- “Find a fiduciary advisor near me who handles RSU/ISO planning.”
- “Best NAPFA advisor for retirement income planning, Phoenix.”
- “Compare flat-fee vs. AUM advisors for a $2M portfolio.”
- “CFP that specializes in widows and inheritance, Atlanta.”
- “Fee-only advisor for tech founders, San Francisco.”
- “Who should I hire to manage my IRA rollover, Houston.”
- “Best advisor for early-retirement planning, Denver.”
- “Tax-aware advisor for sale-of-business proceeds.”
- “Conflict-free retirement advisor for $1M+ portfolio.”
- “Top fee-only CFP for federal employees TSP rollover.”
- “Trusted advisor for special-needs trust planning.”
- “Best advisor for foreign-service or expat compensation.”
Three patterns emerge if you read the list as a whole.
The first is that every prompt names a specialty — physicians, equity comp, RSU/ISO, federal employees, special-needs, expats. The HNW prospect is not asking for a generic advisor. They are asking for an advisor who has solved their specific situation before. AI engines have learned this and now refuse to recommend generalists when the prompt is specific.
The second is that geography is decisive but flexible. “Phoenix,” “Boston,” “Atlanta” anchor the answer to a metro, but engines return out-of-metro firms when the specialty is rare enough — a national-platform RIA that specializes in foreign-service compensation will surface for “expat compensation, anywhere” as readily as for “expat compensation, DC.”
The third is that compliance language matters. “Fee-only,” “fiduciary,” “conflict-free,” “NAPFA” — these are the trust filters HNW buyers apply. AI engines have absorbed the same filters and will preferentially cite firms whose entity-graph signals satisfy them. A firm that does not surface NAPFA membership in machine-readable form on its homepage forfeits the citation on every fiduciary-keyed prompt.
The full 15-prompt cluster — with the specific page-level capsules each prompt rewards, and the directory citations each one pulls from — lives in the RIA ChatGPT citation prompts deep-dive. The prompts there are the operational playbook. The prompts here are the diagnostic — read them and you can predict, prompt by prompt, where your firm currently sits in the citation set.
FINRA Rule 2210 and the SEC Marketing Rule: what GEO must respect
The reason this category has stayed empty is regulatory. Most marketing playbooks rely on the kind of language — performance claims, comparative superiority, testimonial framing — that FINRA Rule 2210 and the SEC Marketing Rule do not allow registered firms to publish. AI visibility content engineered to lift firms into citations has to thread that needle.
FINRA Rule 2210 governs all communications with the public, and the rule does not exempt AI-targeted content. The FINRA 2026 Regulatory Oversight Report, released December 2025, explicitly names GenAI supervision as a focus area for the year. Sidley Austin’s December 2025 client memo on the FINRA 2026 report flagged the practical implication: any content engineered for retrieval by an AI chatbot is communications-with-the-public, and any AI-generated content firms publish or distribute must clear the same principal pre-approval review as any other piece of marketing.
The SEC Marketing Rule layers on a recordkeeping obligation. The rule demands factual accuracy and substantiation for every claim. The Wealth Management magazine review of SEC 2026 examination priorities — published Q1 2026 — confirmed that AI compliance is on the exam list. Firms have to be able to produce, on request, the substantiation for any AI-related capability claim and the audit trail for any AI-generated content.
The translation into engineering terms is short. The compliant GEO surface for an RIA is exactly four layers, and nothing else:
Entity-graph layer. Person + Physician-style structured data — except for advisors it’s Person + FinancialService + hasCredential (CFP, CFA, NAPFA, XYPN). Pure factual data. No claims. The entity graph is what AI engines use to disambiguate a firm and decide the firm exists at the intersection of its specialty + geography + credentials. This layer is invisible to readers and load-bearing for AI citation. The detailed schema map for advisors lives in the FINRA-compliant entity-graph for advisor sites.
Credential and disclosure layer. Static, server-rendered, machine-readable. FINRA-registered representative status; SEC-registered investment adviser status; ADV Form linkage; BrokerCheck linkage; CRD numbers; state registration footprint. Every one of these is factual, every one of them is required disclosure anyway, and every one of them is a citation surface AI engines reward. The compliance review for this layer is short — these are facts, presented as facts.
Bio-and-niche layer. Long-form principal bios that name the specialty (RSU planning, federal-employee TSP rollover, special-needs trust planning) and the credential (NAPFA, XYPN, CFP). No performance claims. No comparative superiority. Just specialty + experience + credential, in long-form prose AI engines can lift verbatim into a citation. This is where the LinkedIn-headline pattern Wealthtender documented translates onto your own site at scale — the same structure, with the citation surface under your own canonical domain.
FAQPage layer. Structured Q&A that surfaces the questions HNW prospects ask the AI before they ask you. “What does fee-only mean?” “What is a fiduciary advisor?” “How does an AUM fee compare to flat-fee planning?” Each one is a question + answer pair, both server-rendered. AI engines lift the answer block when the prompt matches. This works under FINRA because the answers are educational, factual, and free of performance claims. The schema implementation pattern lives in the FAQPage schema and the citation lift pattern.
What this list excludes — by FINRA design — is everything that would normally be the GEO playbook in another vertical. No “best advisor in [city]” landing pages. No comparative “us vs. competitor” content. No client outcome stories without principal review and substantiation files. No performance claims, period. The four-layer surface above is the entire compliant GEO inventory for an SEC-registered investment adviser.
The constraint sounds heavy until you compare it to the alternative: a category most firms have not entered, with a category-leading deal economics, where the compliant playbook is structurally narrower than the AI engines’ citation criteria. The narrowness is the moat.
The platform problem: where most advisor sites cap their own citation ceiling
There is no advisor-specific BuiltWith report for 2026, but the working consensus across advisor-marketing reviews is that the category runs on three template-platform layers, in this approximate order.
WordPress (~42.5% of all websites globally as of April 2026, WPZOOM) is the substrate. Most advisor sites are WordPress underneath. The page builders on top — Elementor, Bricks, Divi, Beaver Builder — determine whether the platform is a citation asset or a citation cap. The detailed comparison lives in the WordPress page-builder GEO ranking.
Advisor-niche template platforms — FMG Suite, Twenty Over Ten, AdvisorWebsites. These run on WordPress underneath but lock the schema-editing surface, lock the canonical-tag editing, and require compliance-archive script injection that often runs client-side. The result is a site whose entity graph is invisible to AI crawlers even when the agency thought they had implemented schema. The diagnostic is one line of curl: fetch the homepage with no JavaScript and grep for application/ld+json. If the entity graph isn’t there, the template is the cap.
Squarespace and Wix Studio for boutique RIAs. A meaningful minority — especially solo CFPs, fee-only planners under $50M AUM, and breakaway-team firms in their first 18 months — built on Squarespace or Wix Studio. Both platforms cap the GEO surface in load-bearing ways. The reason advisor-template platforms cap citation lives in the Squarespace canonical trap — the platform refuses to let you edit your canonical tag, and every site ships with a permanent /home → / mismatch. Wix Studio’s structural problem (the 8K-character schema cap and client-side JSON-LD injection) is the same shape.
The platform layer is decisive for AI citation in a way it is not for Google. Google renders JavaScript and forgives a lot. The AI engines do not render JavaScript and forgive nothing. A site whose entity graph appears only after the FMG Suite injector script runs is a site AI engines do not see structured data on. The technical fix is server-rendered schema in the initial HTML response, which is a constraint most advisor-niche platforms cannot satisfy without migration.
The migration path for a typical $200M-$1B AUM RIA is a static-rendered Astro build with the entity graph in the initial HTML, the four-layer compliant GEO surface above wired in, and the existing compliance-archive integration moved to a server-side recording layer. The build window is 7 to 14 days. The compounding citation graph starts the day the new HTML is live.
What lives in this hub: the advisor playbook clusters
This pillar is the hub for two operational deep-dives. Read them in this order if you are building the advisor playbook from scratch:
The 15 RIA ChatGPT citation prompts, with the page capsules each one rewards. The 15 prompts above, expanded into a per-prompt operational playbook. For each prompt: which directory the engine is currently citing, which page on your own site needs to exist to compete, the answer-capsule format that earns the lift, and the FINRA-compliant phrasing.
The NAPFA, XYPN, SmartAsset, and named-firm directory citation stack. The directory side of the playbook. Per directory: the profile completeness checklist, the entity-graph fields AI engines surface, the linking strategy back to your own site, and the per-state citation share data from the Wave 1 Wednesday data drop.
The cross-hub lateral every advisor playbook should read once is the vertical citation playbooks hub — financial advisors are the Wave 1 vertical because of the deal economics, but the structural pattern (entity-graph + directory stack + compliance-respecting on-page surface) repeats across med-spas, plastic surgeons, dentists, and lawyers. Reading the cross-vertical pattern once compounds the per-vertical work.
The cross-hub for fellow regulated professions: the fractional CFO and SaaS-accounting citation prompt set applies the same FINRA-shaped compliance frame to SEC-regulated accountants and Big-4 alumni. If you serve adjacent client profiles to a fee-only planner, the prompt overlap is high and the entity-graph patterns translate one-to-one.
The Wave 1 data drop: the per-metro citation share scan
The single biggest underserved opportunity in this category is that nobody has measured advisor-specific AI citation share at metro resolution. The 5W AI Platform Citation Source Index is cross-vertical. The Wealthtender and Edge Partners reviews establish that NAPFA, XYPN, LinkedIn, and BrokerCheck are the citation surfaces in aggregate. Nobody has run the prompt — “best fee-only fiduciary advisor in [Austin / Boston / Phoenix / Denver / etc.]” — across the 25 largest US metros and recorded which directories the engines actually cited and at what share.
That data drop runs Wednesdays on this hub. The categories the drop measures, per metro, per major engine (ChatGPT, Gemini, Perplexity, Claude):
- NAPFA Find an Advisor citation rate (% of metro prompts that surface a NAPFA URL)
- XYPN adviser-search citation rate
- SmartAsset SmartAdvisor citation rate
- BrokerCheck and CFP.net citation rates
- LinkedIn profile citation rate, with the headline-pattern that surfaced
- Named-firm citation rate (the firm’s own canonical homepage or bio page)
- Reddit thread citation rate (per the 5W cross-engine pattern, ~3% on ChatGPT specifically, higher across Perplexity)
- Wikipedia citation rate (~5% per-citation, ~18% per-conversation cross-vertical baseline)
The first metro batch publishes the Wednesday after this article goes live. The methodology is documented at the head of the data-drop article so other firms can run the same scan against their own city. The data is the asset; the methodology is the credential.
The reason this matters strategically — beyond being the first study in the category — is that the Wednesday original-data cadence is itself a citation lift. AI engines reward freshness on a 24-to-72-hour retrieval cycle and reward original data above synthesized content. A weekly drop of fresh, citable, methodologically-disclosed metro-level advisor citation data is exactly the content profile the engines preferentially surface in 2026.
The 50× math, restated
A $5M AUM client at a 1% AUM fee returns $50,000 per year, recurring. The site rebuild and entity-graph engineering that gets a fee-only RIA into the citation set is a one-time spend. One client pays it back fifty times over. The AI citation graph compounds — every prompt the firm enters, every directory profile that surfaces, every Wednesday data drop the hub publishes — and the second, third, and fourth clients arrive at the same fixed cost.
The constraint is the platform layer. The advisor-niche template platforms — FMG Suite, Twenty Over Ten, AdvisorWebsites, plus Squarespace and Wix for the boutiques — cap the entity graph behind client-side rendering AI crawlers do not see. The compliance constraint — FINRA Rule 2210 and the SEC Marketing Rule — narrows the on-page surface to four layers (entity graph, credentials, bios, FAQPage). And the empty-category constraint — no public per-vertical citation share data exists — is the moat for whichever firm runs the measurement first.
The buyer behavior is documented. The directory short list is documented. The compliance frame is documented. The platform problem is documented. The only thing left to do is build the site that satisfies all four at once and ship the Wednesday data drop that puts the firm on the citation surface AI engines now reward above any other content profile.
The 1-in-4 HNW buyers using ChatGPT in 2026 are the prospects already there. The shortlist they read tomorrow morning is the shortlist that closes.
