The Black Box Effect, Type 2, is expertise that was never written down at all: the methodology you run but never named, the decision framework that exists only in how you practice, the war stories that come out on client calls and nowhere else. AI assistants can only weigh what has been published, so the version of your business they can recommend is the written-down version. For most experts, that is the weakest version. The brochure, not the brain.
If this is you, the diagnosis will feel almost unfair, because nothing about it involves you having done anything wrong. You built the firm on referrals and results. The work was always in front of you, so the phone kept ringing and the relationship sold itself. Every hour you did not spend writing was an hour spent being good.
The catch is that the referral system that rewarded all of that has a new gatekeeper, one that reads instead of remembers. AI assistants now sit in your buyer’s first conversation, and they can only weigh what got published.
At Probably Genius this is the condition we were most specifically built for, and our published methodology page on the Black Box Effect is the canonical definition of the whole framework. This article is the guide to its second half: why the best practitioners are so often the least documented, what the gap between the business you run and the business the internet describes actually costs, and how the Genius Interview turns twenty years of unwritten judgment into a record the machines can finally quote.
What Lives Only in Your Head
The Black Box Effect, as defined on our methodology page, is genuine expertise that becomes invisible to AI because it lives in formats machines cannot open. Type 2 is its purest form: the expertise was never captured in any format in the first place. It shows up only in direct relationship, in the meeting, on the call, in the moment a client brings you a mess and watches you see through it.
Take a quick inventory of what tends to live there. The methodology you actually follow with every client, which has steps, sequence, and rules of thumb, and no name. The intake questions competitors never think to ask. The pattern you spot in the first ten minutes because you have seen this exact situation forty times.
Add the contrarian positions you hold with total confidence and have only ever defended out loud. Add the war stories, the near-disasters and saves that would teach a buyer more about your judgment than any credentials page, told at dinners and never typed. For most veteran practitioners, that inventory is the entire differentiating layer of the business. And none of it exists, in the only sense the machines understand existence.
The methodology page names the cause honestly: the relationship-selling trap. For decades, referral-driven firms had no commercial reason to write anything down. Knowledge moved person to person, reputation moved mouth to mouth, and the work spoke for itself to everyone who saw it. As the page puts it, “For 20 years, the relationship sold itself. Now you need to build a relationship with AI — so it can sell you when you’re not in the room.”
The trap was never laziness. The business model made writing things down genuinely unnecessary, right up until the year it became the whole ballgame. And be precise about who this describes, because it is not the beginner. Type 2 is a senior practitioner’s condition; it takes years of practice to accumulate a methodology worth trapping. Which sets up the cruelest feature of the whole effect.
The Trenches Problem: Why the Best Are the Least Documented
Here is the irony at the center of Type 2, and once you see it you will see it everywhere: the reason these experts are invisible is the same reason they are good. Our co-founder Jacquie Baker describes the people this happens to in the language of where they actually spent their careers: they “were in the trenches figuring out real-life applications instead of sitting in glass palaces intellectualizing concepts to get published.”
The ones with the deepest expertise were too busy serving clients, solving problems, and building to stop and write about it. They were in the trenches. The people AI cites were getting published.
This is not a new observation about knowledge. It is a new cost attached to a very old one. In 1966 the philosopher Michael Polanyi published The Tacit Dimension and gave the phenomenon its permanent name, starting from the fact that “we can know more than we can tell.” Every craft and profession has known masters who could do what they could not fully explain.
For sixty years that was a philosophical curiosity and a knowledge-management headache. Then a system arrived that sits between experts and their future clients and can only read what got told. Polanyi’s gap stopped being academic the day the referral started going through a machine.
And the machines are explicit that lived experience is what they are looking for. In December 2022, Google added Experience to its search quality rater guidelines as a named dimension alongside Expertise, Authoritativeness, and Trustworthiness: does the content show genuine first-hand experience of the topic? Semrush’s January 2026 study of 11,882 prompts across ChatGPT Search, Google AI Mode, and Perplexity found E-E-A-T signals among the traits most correlated with being cited, at +30.64 percent.
Sit with the full unfairness of that for a moment. The systems reward demonstrated experience in content, and the people with the most experience produced the least content. The evaluation framework reads like a portrait of the practitioner in the trenches, and it can only ever score the practitioner’s paper trail.
The Two Versions of Your Business
Every established firm now has two versions of itself. There is the business you run: the unnamed methodology, the pattern recognition, the twenty years of saves. And there is the business the internet describes: a services list, a team page, some testimonials, maybe a blog of generic explainers an agency produced years ago.
Buyers who reached you through referrals got introduced to the first version, with a trusted friend’s endorsement carrying all the undocumented depth. An AI assistant answering your buyer’s question has only ever met the second.
That swap happens silently, which is what makes it expensive. When a buyer asks who can handle a complex, high-stakes problem, the engine weighs identities it can verify, claims it can corroborate, and answers it can quote. Your firm’s entry in that weighing is the brochure version. It competes against firms whose written record is thick with named methods and demonstrated judgment, some of whom are, in the trenches sense, half as good as you.
The machine is not being fooled. It is doing exactly what it is built to do with exactly what you gave it. Jacquie’s line from her January 2026 essay The Silent Vetting is the whole section in two sentences: “Most experts aren’t invisible because they lack expertise. They’re invisible because their expertise is trapped.”
There is a layer above ordinary search where this verdict gets rendered, where the machines stop listing options and start giving answers. We call it the Recommendation Layer, and a place in it is decided on the written record. So the gap between your two versions is not a branding annoyance. It is the exact measure of how much earned authority you are leaving out of the decision.
Want to see your own gap? The test is short: ask an AI assistant your buyers’ real questions, in plain sentences, and read which version of you answers back. Our plain-English guide to what AI visibility is walks through exactly how to run it.
How the Genius Interview Works
You cannot transcribe what was never recorded, so Type 2 has a different fix from Type 1. Ours is called the Genius Interview: a structured founder interview built to get the unwritten layer out of your head and onto the record, in your voice, under your name. It is not a marketing questionnaire, and it is nothing like being handed a blog calendar. It is closer to a deposition taken by someone on your side.
The craft is in what gets asked. The interview locks the anchor facts first: the names, places, specialties, and boundaries that define who you are and who you are not for. It walks your actual process step by step until the unnamed methodology has edges, a sequence, and finally a name.
It hunts the war stories deliberately, because a specific save demonstrates judgment the way no adjective can. It asks for the cost of inaction in real numbers. And it asks the question that reliably produces the material nothing else reaches: why you do this, and not the business reason.
What comes back gets mapped into your canonized IP: the methodology named and defined, the decision frameworks stated as claims, the vocabulary that is yours, the stories attached to the judgments they prove.
Two things about this matter more than the mechanics. First, the expertise is already finished. Nothing gets invented, embellished, or ghost-thought; you have been rehearsing this material on client calls for twenty years, and the interview simply captures a performance you have already perfected. You are the genius in this arrangement. Our whole job is translation, from the room where you say it to the record where machines can read it.
Second, everything gets confirmed with you before it is published, because the point is a record that is finally accurate, not merely finally loud. Jacquie describes the experience the way clients actually receive it: “my whole career is holding a mirror back at you, showing the beauty I see — and this is what AI is.” The interview does not add anything to you. It shows the machines what was already there.
The Content AI Has No Other Way to Get
Here is the strategic payoff, and it is bigger than filling a gap. The engines are drowning in content that says what every other page already says, and they have visibly shrinking reasons to cite any single copy of it. What they lack is material that exists nowhere else: named methods, first-hand numbers, positions with an author attached.
Mapping what has never been published produces exactly that, which is why we call the output Net New Intelligence. It is net new by definition. A methodology mapped out of one practitioner’s head cannot be a photocopy of anything, and it hands an engine something it structurally cannot get from anyone else in your category: an original source.
This is also where the work compounds instead of just accumulating. Once the methodology has a name and a canonical page, every future article, mention, and citation can point at it, and the machine’s file on you, the entity work we explain in plain English in our entity identity methodology, gains a center of gravity. The war stories become demonstrations of the named method. The contrarian positions become claimed territory.
Ask the engines about the problem you solve, and there is finally a specific, verifiable, quotable body of work with your name on it standing where the brochure used to stand. That is the difference between being findable and being the answer. For the expert whose genius spent twenty years locked in the room, it is overdue justice: the machines rewarding you, at last, for the thing you were actually best at.
Want to Learn More?
Probably Genius was built by Jacquie Baker and Christopher Shaw, who spent more than 40 combined years translating what makes an expert the best in the room for the audiences that decide who gets chosen. We map genius, build the infrastructure that makes it verifiable, and publish the proof that gets it recommended.