The AI Visibility Gap

Score your own firm

Before the detail, a gut check. This study scored 130 firms on five dimensions of machine-readable identity. Answer five plain questions and see where your own firm would likely land. Not knowing an answer is itself the finding, because for most firms the work was never done. The sections below explain exactly what each answer means.

The findings are technical. What they mean for a firm that depends on being found is not.

How your prospects now search

They no longer scan ten blue links. They read one AI-generated summary and act on it. Pew Research tracked the actual browsing of 900 US adults in March 2025: fifty-eight percent ran at least one search that returned an AI summary, and when that summary appeared, they rarely clicked through to the sources behind it (Pew Research Center, 2025a; 2025b). The summary is the answer now.

And the people who trust it most are your prospects. Sixty-three percent of upper-income Americans report at least some trust in AI summaries, against fifty percent of lower-income (Pew Research Center, 2025c). The high-net-worth households, the founders pre-liquidity, the families choosing a next custodian: that is the cohort most inclined to take the AI's answer at face value. Among adults under 30, the generation now inheriting that wealth, half use AI tools weekly (Brookings, 2026).

So if the summary leaves you out, or lists you generically, the prospect never reaches your site. Your homepage is no longer the first thing they read. The summary is, and it is built without your input.

Why you are not in the summary

An AI assistant does not read your site to learn who you are. It synthesizes from sources it has already indexed and learned to trust. The sources that surface in advisor answers, per Kitces.com research, are external to you: Reddit, YouTube, SmartAsset, Wealthtender, the Fee-Only Network, NAPFA (Kitces.com, 2025). Not your blog. Not your about page. Not your principal's bio.

That is the whole problem. The AI reads what the directories said about you, and the directories describe the entire category identically — fee-only, fiduciary, CFP-led. Your fifteen years in equity compensation, your liquidity-event specialization, may sit right there on your site and remain invisible to the layer doing the synthesizing.

The reason is that AI rewards verified signals. It cannot tell a true claim from a marketing claim by reading prose, only by checking whether the claim is anchored: schema linking the firm to its SEC registration, the principal to their CFP, the article to its author. Schema is simply the structured form your information takes so a machine can verify it. The plain analogy: it is signing your name and giving the address where the AI can confirm everything you signed. Without it, the AI reads the signature and has nowhere to check it.

Fifty-nine percent of audited firms have signed nothing at all. That is the gap. It is also closeable.

The 130 firms in this study form the analysis base of a 150-firm sample selected through stratified random sampling from the IAPD bulk XML feed (IA_FIRM_SEC_Feed_05_26_2026.xml.gz, 23,395 total registrants). After filtering to full registrants with approved status, a US principal office, and AUM of at least $25M, the eligible population was 13,953 firms, of which 13,089 carry a functional website URL on Form ADV. The 150-firm sample represents approximately 1.1% of the eligible auditable population. Eligibility for the consumer-facing analysis was restricted to firms with a public-facing website directing services to individual investors or families, and no institutional fund or B2B advisor platform structure.

Stratification targeted four simultaneous dimensions. Geography stratified across Tier 1 metropolitan markets (40 percent), Tier 2 metropolitan markets (35 percent), smaller markets (20 percent), and rural markets (5 percent). Firm size by assets under management stratified across solo and small firms at $25M–$250M (30 percent), small team firms at $250M–$1B (30 percent), mid-size firms at $1B–$5B (25 percent), large firms at $5B–$50B (12 percent), and mega firms above $50B (3 percent). Target market and compensation structure (fee-only at 50 percent, fee-based or hybrid at 35 percent, commission at 15 percent) were also stratified. Random seed 20260526 is documented for reproducibility.

Each firm was audited across four sampled pages: homepage, about page, one service page, and one article or blog post where present. Schema data was extracted via live JavaScript execution on each page rather than from raw HTML, ensuring that client-side-rendered schema and dynamically-injected JSON-LD blocks were captured. AI visibility was measured using Perplexity with two queries per firm: a brand-name recognition query and a category-based search query for the firm's market and geography. Both queries were run in authenticated sessions with no prior conversational context. Claude, the model that conducted the audit, was excluded from AI-visibility testing as a conflict-of-interest measure, so that no model evaluated outputs from its own class.

Of the 150 firms drawn, 20 were classified as not applicable after review: private equity and hedge fund managers, sub-advisory platforms serving institutions rather than individuals, credit unions and fund administrators, and hospital-system fiduciaries with no public-facing advisory practice. These 20 firms are excluded from all findings. The 130 consumer-facing firms form the analysis base. The full eligibility definition and exclusion rationale are documented in the study's methodology file.

The five dimensions audited were scored on a one-to-five scale per firm. Schema coverage measured presence, type-correctness, and propagation of structured data. Entity coherence measured the consistency of identifier references across pages and the integrity of the firm's entity graph. E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness) signaling measured the presence of Person entities, credential declarations, and article author schema. External verification measured the presence of sameAs links to authoritative external sources: regulatory records, professional networks, business profiles. AI visibility measured how confidently and specifically the firm was surfaced by an AI search assistant in direct and category-based queries.

Eighty-nine percent of audited RIA firms surface in AI search, scoring four or five out of five on AI visibility. Fifty-nine percent of those same firms have no structured organization entity anywhere on their website. Sixty-nine percent score AI visibility of four or five while their schema coverage scores two or below. They are visible despite the absence of meaningful schema infrastructure.

The central paradox of this dataset: advisory firms are found by AI at a rate their own infrastructure does not justify. They are indexed by third-party directories — IAPD, SmartAsset, FinanceStrategists, Advisor Pages, WiserAdvisor, Bankrate — which scrape regulatory data and present it in forms AI can read. Kitces.com research finds the same pattern, naming Reddit, YouTube, SmartAsset, Wealthtender, the Fee-Only Network, and NAPFA as the consistent surfacing points (Kitces.com, 2025). None of these is the firm's own site.

The gap between AI visibility (mean 3.81/5) and schema infrastructure (mean 1.93/5) is 1.88 points. That is not a curiosity; it is a window, closing on a timescale of quarters as AI search matures toward verification.

What this means. Firms whose visibility lives entirely in third-party directories are renting it from infrastructure they do not control. The directory can change its structure, deprecate its feed, or update less often, and the firm has no recourse. Firms anchored in their own schema own that visibility outright. The first to make the switch widens the gap behind them.

The mean schema coverage score across the 130 firms is 1.93 out of 5.00. Sixty-two percent of firms have JSON-LD on at least one audited page, a deceptively encouraging number, because the presence of structured data is not the same as the presence of useful structured data. A WebPage schema block carrying a bare URL is metadata about a document, not a declaration of who the firm is.

Structured data investment across the independent RIA space is, with narrow exceptions, either absent or badly deployed. The specific breakdown is as follows.

The most consequential gap is not the absence of JSON-LD. It is the absence of any organization entity. Fifty-nine percent of firms have no Organization-type entity declared anywhere, meaning no Organization, FinancialService, or LocalBusiness entity exists as the anchor for the firm's identity. From a knowledge-graph perspective, such a firm is a collection of documents with no identified author. Of the 41 percent that do declare an organization entity, only 15.4 percent of the full sample use the semantically correct FinancialService type; the rest use generic Organization or LocalBusiness, which fail to signal to AI parsers that this is a regulated financial professional. Only 17.7 percent maintain a stable @id propagating across pages, the foundational step for entity resolution across multiple URLs.

Among the fifteen firms with FinancialService schema declared, the most common failure mode is the same across firms: the correct type is present, but the connective tissue is missing. No @id anchor binds the entity across pages. No sameAs links extend the entity to external authoritative sources. A declaration of identity exists without an actual identifier — like signing a name without giving an address.

What this means. The independent RIA category does not have a schema problem in the sense of firms that built schema infrastructure incorrectly. It has a schema problem in the sense of firms that never built it at all. The remediation work for any individual firm is therefore not refactoring or correction. It is initial construction. The first firm in any given market or niche that completes that initial construction will be the firm AI search systems can verify and recommend by name, while competitors continue to appear as directory listings.

Zero of 130 firms link to their IAPD regulatory record in their structured data. Zero of 130 fee-only firms link to their NAPFA membership in their structured data. Eighty-two percent carry no authoritative external links in schema at all.

Every firm in this study is registered with the Securities and Exchange Commission as an investment adviser. The regulatory record for each firm is publicly accessible via the IAPD database at adviserinfo.sec.gov. This record includes CRD number, Form ADV disclosures, registration history, disciplinary actions if any, and principal officer names. This is precisely the kind of authoritative external verification that entity graphs are designed to link to.

Not one firm in the sample has made that connection in its structured data.

For AI systems attempting to verify a firm's claimed credentials (SEC registration, principal's CFP designation, absence of disciplinary history), the absence of this link means the verification must happen indirectly. The model must trust scraped directory data, which itself derives from IAPD but at one or more removes from the firm's own assertion. The regulatory authority layer that would make advisory firm AI visibility durable is not just rare. It is categorically absent.

The same observation holds for NAPFA membership. Fee-only status is a primary differentiator in the advisory market, claimed prominently by virtually every fee-only firm in this study, often as the central element of the firm's positioning. Yet no firm has asserted that status in schema by linking to its NAPFA directory entry, which would allow AI systems to cross-reference the claim against an independent authoritative source. The same applies to FINRA BrokerCheck for firms with FINRA-licensed principals.

What this means. The regulatory layer is the most expensive thing to fake and the cheapest thing to verify — and nobody has built it. The records sit at stable, public, CRD-keyed URLs (brokercheck.finra.org/firm/summary/[CRD]), and schema.org's sameAs property exists precisely to hold them. For firms with FINRA-licensed principals the case is stronger still: FINRA Rule 2210 already requires a visible BrokerCheck link on the site, so adding it to schema extends an existing obligation rather than creating new exposure. This is not a feasibility barrier; it is an execution gap. The first firm to link its data to IAPD, BrokerCheck, and NAPFA moves from being identified to being verified — a signal no competitor without it can match.

The mean E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness) signal strength across the 130 firms is 1.07 out of 5.00, the lowest scoring dimension in the study. Ninety-five percent of firms score the lowest possible value of one.

Google's E-E-A-T framework has a direct structural equivalent in schema markup. Person entities represent named advisors. sameAs links connect those Person entities to their professional profiles. hasCredential declarations document earned designations. Article schema with author references binds bylined content to the named author.

The advisory profession is one in which the credentials of the humans providing advice are the central value proposition. The CFP, CFA, CPA/PFS, and JD designations advisors have spent years earning represent substantial professional authority. None of that authority is represented in structured data for roughly ninety-seven percent of the firms in this study.

Only four firms in the entire 130-firm sample (3.1 percent) have a Person entity identifying the principal advisor. Almost none have article author schema backing bylined content. Not a single firm in the sample declares principal advisor credentials via hasCredential in schema.

This absence has a compounding effect. Advisory firms in this sample publish content at a meaningful rate: market commentaries, financial planning guides, retirement planning articles, equity compensation advisories. That content is their primary vehicle for building trusted expertise online and the central asset that distinguishes a thoughtful independent advisor from a generic firm with the same generic services menu. Without author schema backing that content, the expertise the content represents is invisible to the structured-data layer AI systems use to evaluate source authority. The advisor's CFA is unrelated, in the model's reading, to the article they wrote about portfolio construction.

What this means. The principal advisor is the brand of an owner-operated independent RIA. The schema infrastructure that makes the principal legible includes Person entity, credential declarations, professional sameAs links, and byline-to-author binding. Together these are the single most important infrastructure a firm can build. They make generic content authority-bearing and connect decades of professional credentialing to structured signal. Three percent of audited firms have started on this work. Ninety-seven percent have not.

Firms on the FMG Suite platform have a mean schema coverage of 3.83 out of 5.00. Every other platform in the study combined has a mean schema coverage of 1.74 out of 5.00. The FMG Suite advantage over the rest of the market is +2.10 points.

FMG Suite is a financial-services web platform that injects FinancialService schema by default on every page of the sites it powers. The impact on schema scores is dramatic and consistent, and it is not attributable to the firms themselves. The platform decision, not the firm's digital strategy, determines whether the entity is declared.

FMG Suite firms average 3.83/5 on schema coverage. Every other platform in the study combined averages 1.74/5. The platform gap of 2.10 points is approximately the difference between has no schema and has FinancialService properly deployed.

This finding has two distinct implications.

First, the schema infrastructure problem across the advisory industry is largely a platform problem, not an advisor awareness problem. Most advisors do not know what JSON-LD is. The platform their site runs on either handles structured data or does not. WordPress with default themes generally does not. Squarespace generally does not. Custom builds generally do not unless the developer was specifically asked. FMG Suite does. That platform decision, often made years before AI search became a consideration, now determines a firm's baseline schema position.

Second, FMG Suite's default schema is itself incomplete. It injects FinancialService with social sameAs links (Twitter, LinkedIn, Facebook) but never with the IAPD regulatory link that would make that schema authoritative. The best platform baseline in the advisory market is still missing its most important verification element. For firms on FMG Suite, the infrastructure problem is not absent baseline; it is absent regulatory verification layer above the baseline. For firms not on FMG Suite, the problem is more fundamental.

What this means. The FMG Suite firms in this sample are positioned advantageously relative to the rest of the field. They start from real schema infrastructure and need to build the differentiation layer above it: IAPD, NAPFA, Person entities, credential declarations. The other ninety-one percent need to build from scratch. Both groups have substantial remediation work, but the work is qualitatively different and the cost-of-inaction is asymmetric. A non-FMG firm not just behind FMG firms; it is behind the directories indexing it.

Eighty-nine percent of audited firms surface in AI search, scoring four or five out of five on AI visibility. Six percent score one or two, meaning Perplexity cited the firm's own website rather than a directory, the higher-quality outcome. The mean AI visibility score is 3.81 out of 5.00.

A score of four or five on this dimension means Perplexity surfaces the firm in response to a brand-name or category query, but does so by citing a third-party directory (IAPD, FinanceStrategists, SmartAsset, Advisor Pages, WiserAdvisor, Bankrate) rather than the firm's own website. The firm is visible in AI, but the AI is not reading the firm's content. It is reading a directory that scraped the firm. The 89 percent visibility rate is driven primarily by third-party advisor directory indexing, specifically the sources catalogued in Finding 1.

Only eight firms in the sample (6.2 percent) scored one or two, meaning Perplexity cited the firm's own website as a primary or significant source rather than a directory. Several of these are FMG Suite firms with FinancialService schema deployed; others have accumulated enough content authority that AI systems treat them as primary sources. The causal relationship is not proven by this study, and the subgroup is small. But the pattern is consistent. The firms whose own sites get cited are disproportionately the firms with the best structured data. The remaining ninety-four percent exist in AI search exclusively through third-party mediation.

The fragility of the 89 percent visibility rate lies in its source. When Perplexity answers a query like "Tell me about [Firm Name]," it draws on data indexed from databases that Clarion has no influence over, that update on their own schedules, and that may not reflect firm changes in real time. The firm that moves, rebrands, hires a new principal, or updates its services will see those changes reflected in third-party directories weeks or months later, if at all. The firm will see those changes reflected in its own schema in the same deployment session it makes them.

What this means. Current AI visibility for the independent RIA category is real, but it is borrowed. It depends on the continued presence and quality of third-party directory data. AI search systems are maturing toward weighting authoritative first-party data more heavily. As they do, the gap between firms with their own infrastructure and firms borrowing from directories will widen quickly. The current 89 percent baseline is not a defensive position. It is a starting point that erodes if untreated.

Reading across the findings, three patterns recur with enough consistency to deserve naming.

Pattern one: visibility is borrowed, infrastructure is missing. Eighty-nine percent of firms are legible to AI only because directories did the work — without firm input, verification, or control. The same firms built almost nothing of their own. Recognition is present; ownership is not. This holds across size, geography, AUM, and compensation structure. It is the industry's posture, not a sub-segment.

Pattern two: the regulatory verification layer is missing across the board. Zero firms link IAPD. Zero link NAPFA. Zero link BrokerCheck. The most striking finding in the study, because it is so uniform and so cheap to fix: the records are public, the URL pattern documented and stable, and still nobody has done it.

Pattern three: the human expertise is invisible to the layer that matters. Ninety-seven percent have no Person entity for their principal. None declare credentials in schema. The advisor's training, designations, and authorship — the actual reason a client chooses one firm over another — are nowhere the AI can read. In an owner-operated firm the principal is the brand, and the principal is absent from the data.

Three angles on one finding: the category is recognized as a class but undifferentiated as members, named but unverified, represented as legal entities but severed from the human expertise that is their whole value.

Best practice for independent RIA digital infrastructure has emerged from the data in this study not as theoretical recommendation but as empirical contrast: what the highest-scoring firms have begun to build and what the rest of the field has not. A fully constructed entity graph for an independent RIA firm includes the following elements. No firm in this 130-firm study has all of them in place.

The Organization layer. A FinancialService schema entity, declared in JSON-LD, deployed on every page of the site, anchored by a stable HTTPS identifier (@id) that propagates across all URLs. The @id should be the firm's primary canonical URL with a fragment identifier (for example, https://example-advisors.com/#organization). The entity should declare its name, legalName (matching IAPD registration), url, logo, address, telephone, and description. It should include sameAs links to every authoritative external source where the firm appears.

The regulatory verification layer. Within the FinancialService entity's sameAs array: the firm's IAPD record (https://adviserinfo.sec.gov/firm/summary/[CRD]), the firm's NAPFA directory entry where applicable (https://www.napfa.org/financial-planning/[firm-name]), the firm's FINRA BrokerCheck entry where the firm holds FINRA licensure, the firm's Google Business Profile, and the firm's LinkedIn company page. Each of these is a public URL the AI system can dereference to cross-validate the firm's claimed identity. Zero firms in this study have built this layer.

The Person layer. A Person schema entity for each principal advisor, declared with stable @id, name, jobTitle, worksFor (referencing the firm's @id), and sameAs links to the principal's individual IAPD record (https://adviserinfo.sec.gov/individual/summary/[CRD]), their LinkedIn personal profile, and any professional bio pages where they appear. Three percent of firms in this study have begun this work.

The credential layer. Within each Person entity, hasCredential declarations using EducationalOccupationalCredential schema for each earned designation: CFP, CFA, CPA/PFS, JD, ChFC, CPWA, CIMA, and others. Each credential should declare its credentialCategory, recognizedBy (the credentialing body), and dateCreated where known. Zero firms in this study have made any hasCredential declarations in schema.

The content layer. Article (or BlogPosting) schema on every published piece of content, with the author field referencing the principal's Person entity @id. This binds bylined content to the named author and surfaces the author's credentials in the same entity graph as the article. Almost no firms in this study have implemented article author schema.

The breadcrumb and navigation layer. BreadcrumbList schema on subordinate pages, declaring the path from the homepage to the current page. This is mechanical, schema-native, and important for AI search context.

The discovery file layer. A canonical sitemap.xml listing all indexed pages. A robots.txt welcoming the AI crawlers the firm wishes to be read by (PerplexityBot, ChatGPT-User, GPTBot, Google-Extended, ClaudeBot, Applebot-Extended, and others). An llms.txt file listing the firm's primary content for AI consumption. A security.txt file at /.well-known/security.txt. These are the files that increasingly determine what AI crawlers read and how they weight what they read.

The propagation principle. Every element above must propagate consistently across the four page types audited in this study: homepage, about page, service pages, article pages. Schema deployed on the homepage alone is incomplete; schema deployed inconsistently across pages causes entity disambiguation failure. Propagation is the difference between a firm with schema and a firm with a verified entity graph.

The work required to build all seven layers is substantial but bounded. Each layer is well-defined, the required schema vocabulary is standardized, and the external records to link against are public and stable. The compounding return is durable AI search visibility anchored in the firm's own authoritative infrastructure rather than in third-party directory inference.

Three conclusions emerge.

The category has no baseline of digital authority infrastructure. This is not a problem of firms that built schema wrong. It is firms that never built it. So the opportunity for any one firm is not marginal improvement; it is categorical differentiation.

The first-mover window is open and closing. Today's 89 percent visibility is borrowed from directories. As AI search systems mature beyond directory scraping toward entity verification, the firms that have built durable infrastructure will maintain and sharpen their visibility. The firms relying on third-party directory data will fade into a long, undifferentiated tail. Consumer adoption says this is measured in quarters, not years: AI summaries now reach a majority of US searchers (Pew Research Center, 2025a; 2025b), and half of adults under thirty use AI tools weekly (Brookings, 2026). This is established consumer infrastructure, not an experiment.

Platform choice is determinative, but not destiny. The FMG Suite firms have a better baseline only because the platform handed it to them. Everyone else — WordPress, Squarespace, Wix, custom — got no schema by default and never added it. Proper infrastructure takes deliberate construction, most reliably on a purpose-built foundation. The firms that prioritize it will pass even the FMG baseline by adding the regulatory and Person layers FMG never provides.

The next decade of an independent RIA's competitive position turns on how legibly its infrastructure reads to the AI systems serious prospects now use. Build the entity graph this report specifies, and you are findable, verifiable, and recommended for the specific reasons that match a specific client. Skip it, and you remain a generic listing among many — found, but never chosen.

That is what the AI Visibility Gap costs in 2026. It is closeable, and almost no one has closed it yet.