There is a way to find out exactly how an AI assistant decides which advisor to recommend. You ask it.
Open Claude or ChatGPT and type: "If someone asked you to help them find a financial advisor, walk me through exactly what you would do." The answer is not evasive. The models describe their own process in plain language, and the description is more useful than most of what has been written about AI search optimization.
Here is the part that matters. Asked this question directly in a published interview, Claude explained that it has no database of advisors, cannot browse a directory in real time, and starts instead with a web search, then reads the pages that come back to find the most credible, specific, relevant match to the person's stated situation (Theder, 2026). Its own example: an advisor who has written several pieces on Social Security timing for federal employees will surface when someone asks about exactly that, because the match is specific and the evidence is visible. An advisor who describes themselves as a fee-only fiduciary serving high-net-worth families will not surface for anything in particular, because that description fits ten thousand firms equally well.
That is the whole mechanism, stated by the mechanism itself. The AI is not ranking advisors by quality. It is matching a specific need to the most specific, most verifiable answer it can find. Specificity and verification are the currency. Almost no independent RIA is paying in it.
The reader's own test
Before going further, run the test the AI itself implies. Open ChatGPT or Perplexity and ask it the question your ideal client would ask. Not "find me a financial advisor" — the real one. "I'm a physician in [your city] with a practice sale coming and concentrated stock; what kind of advisor should I look for, and who fits?"
Read what comes back. If your firm is not named, that is not a verdict on your competence. It is information about your legibility. The two are not the same thing, and the gap between them is the entire subject of this essay.
What the AI is reading when it reads you
When an AI assistant evaluates an advisory firm, it is not reading the firm's website the way a prospect would. It is not moved by the photography, the tagline, or the carefully worded value proposition. It reads for signals it can verify, and it weights what it can confirm far above what it is merely told.
There are four of these signals, and they are worth naming precisely, because the difference between a firm the AI can recommend and a firm it cannot comes down to whether these exist in a form a machine can read.
The first is a specific, named specialization. Not "comprehensive wealth management." A named who, a named what, a named where. The advisor who works with airline pilots navigating early retirement and concentrated company stock is legible. The advisor who serves "individuals and families seeking financial peace of mind" is not. The AI cannot match a generality to a specific query, because a generality matches every query equally, which is to say it matches none of them well enough to name.
The second is content depth around that specialization. One page claiming expertise is a claim. A dozen pieces working through the actual problems of that niche, the tax mechanics, the timing decisions, the mistakes people make, is evidence. The AI treats sustained, specific writing as a signal of genuine expertise in a way it does not treat a services menu. This is the closest thing to a durable moat available to an independent firm, and it is the thing most firms have least of.
The third is a footprint beyond the firm's own walls. The AI trusts a firm more when the firm's existence is corroborated by sources the firm does not control: a current Google Business Profile, third-party reviews, mentions in places the firm did not write. Reviews carry unusual weight here, because a client who writes "finally made sense of my equity compensation" describes a real problem the advisor solved, in words the advisor did not choose. That is corroboration a services menu can never supply. A claim a firm makes about itself is weaker, to a verification-seeking system, than the same fact confirmed somewhere independent. This is why the directories win: they are the corroboration, and right now they are the only corroboration most firms have.
The fourth is the one almost nobody has built, and it is the most decisive: verifiable identity. The structured connection between the firm as it appears online and the firm as it exists in authoritative records: its SEC registration, its principal's credentials, its regulatory standing. This is the difference between a firm that says it is a registered fiduciary and a firm an AI can confirm is one. The first is marketing. The second is verified fact. AI systems, operating under the high-stakes standards that govern financial information, increasingly weight the second and discount the first.
Why almost no firm passes its own test
This is where the data turns the demonstration into an indictment.
Clarion's 2026 audit of 130 independent RIAs measured exactly the signals the AI says it reads. The results, against the AI's own stated criteria, are close to total failure. Fifty-nine percent of firms have no structured identity on their own website at all, nothing a machine can read as this is who we are. Ninety-seven percent have no structured representation of the principal advisor, the person whose name and credentials are the entire basis of an owner-operated firm. And not one firm in the 130, zero, has linked its own structured data to its SEC registration, the single cleanest piece of verifiable identity available to any advisory firm in the country.
Read those numbers against the four signals and the picture resolves. The specialization is often there in the prose, invisible in the structure. The content depth, where it exists, is unattributed to any verifiable author. The external footprint is borrowed entirely from directories. And the verifiable identity layer, the one the AI weights most heavily, is absent across the entire sample.
The firms are not failing because they lack expertise. Most of them have exactly the expertise their prospects are looking for. They are failing because that expertise was written for human readers in an era when human readers were the only audience, and the new reader, the one standing between the firm and the client, cannot see it.
The part you cannot write your way out of
Here is the conclusion that changes what a firm should do about this, and it is the reason a better About page will not save anyone.
If the problem were that the firm's prose was insufficiently compelling, the fix would be to write better prose. It is not. The AI is not reading prose for persuasion. It is reading structure for verification. A firm can have the most eloquent description of its equity-compensation practice ever written, and if that description is not anchored, to a named structured author, to a verifiable credential, to an authoritative external record, the AI reads it the way it reads every other firm's eloquent description, which is to say as an unverifiable claim.
This is why the work is structural, not cosmetic. The signals the AI reads are not tones of voice. They are machine-readable facts: an organization entity that declares what the firm is, a person entity that declares who runs it, credential declarations that connect the principal to their actual designations, a verification link to the regulatory record that confirms all of it. These either exist in a firm's infrastructure or they do not. No amount of rewriting the homepage creates them, just as no amount of describing a building makes it structurally sound.
The independent RIA's instinct, faced with an invisibility problem, is to reach for words: a new tagline, a sharper value proposition, a refreshed About page. It is the wrong instinct for this problem, because this problem is not made of words. The AI already has the firm's words. What it does not have is anything it can verify them against.
What this means for the firm that moves first
The same audit that documents near-total failure documents the opportunity inside it. A signal that no competitor has built is, by definition, available. The first firm in a given market or niche to make its specialization structured, its principal a verifiable entity, and its identity confirmable against the regulatory record becomes the firm the AI can recommend by name — while every competitor remains a generic listing the AI describes but cannot distinguish.
This advantage compounds, because the firms that build it now establish themselves as the verifiable answer before the field catches on, and verification, once built, holds. As AI search systems mature beyond directory scraping toward genuine entity verification, the gap between the firms that did this and the firms that did not will widen, not narrow.
The AI will tell you, if you ask it, exactly what it needs to recommend you. It has been telling anyone who asks. The firms that listen to the answer, and understand that the answer describes infrastructure, not copy, are the ones it will be naming a year from now.