A private diagnostic that measures how extracted, documented, and protected your expertise really is, and maps exactly what needs to happen next.
Before anything gets built, we find out what is already there. The Expertise Audit is the diagnostic that opens every Legacy Building engagement, and the Legacy Score is the number it produces.
The Legacy Score answers one question, and it is deliberately an uncomfortable one: if you stopped creating new content today, how much of your expertise would still be discoverable and citable twelve months from now?
Most experts have never been asked this. They measure output — posts published, episodes recorded, followers gained. Output is a measure of effort. The Legacy Score is a measure of what survives the effort.
For most of the 2000s and early 2010s, more content meant more surface area, and more surface area meant more visibility. That logic held until retrieval changed. The systems that now answer your buyers' questions — AI Overviews, ChatGPT, Perplexity, Gemini, Claude — do not rank a library. They pull a passage.
Those systems do not reward volume. They reward citability: being specific, structured and authoritative enough that a machine assembling an answer from millions of sources chooses your formulation over someone else's.
The research is unusually consistent about what that means in practice. Adding named statistics and cited sources to a passage lifts its likelihood of being surfaced by roughly 30–40%. Tables are pulled meaningfully more often than the same information written as prose. Roughly 88% of citations in Google's AI Mode come from pages outside the organic top ten — meaning your search ranking is close to useless as a predictor of whether an AI will quote you. And the overwhelming majority of what these systems cite is earned: independent sources describing your work, not your own marketing describing it.
None of that is fixed by publishing more. It is fixed by publishing differently, and by knowing precisely which parts of what you already know are worth the effort.
Before you can audit what you have, you need a vocabulary for what you are looking at. Almost all professional content falls into one of three categories. Only one of them builds authority that lasts.
Content only you could have produced: a framework built through years of client work, a distinction you invented to name a problem nobody had named, a mechanism in your domain that explains why interventions succeed or fail. It cannot be replicated, because it reflects one person's accumulated pattern recognition. This is what AI systems treat as an authority source — it exists at higher concentration nowhere else.
Accurate, useful, well-written — and already available in ten thousand other places. Best-practice roundups, industry news commentary, explanations of things the field already agrees on. It builds trust with a human reader who is already listening. It contributes almost nothing to machine-legible authority, because a retrieval system has no reason to prefer your version.
Reactions, hot takes, comment threads, stories, posts on platforms you rent rather than own. It performs while it is live and disappears from the retrievable record almost immediately. It is not wasted — it is how audiences are built — but it should never be mistaken for a deposit into the archive.
The Framework
Each stage builds on the one before it, moving from what you have already published, to what exists only in your head, to the distance between the two — and finally to a number you can act on.
Everything you have already put into the world, gathered in one place and classified honestly against the three categories above. Articles, episodes, decks, course modules, book chapters, long client emails, conference talks. Not counted — classified. Most experts discover that a large body of work contains a surprisingly small amount of proprietary signal.
Ask: of everything I have published, which pieces could only have come from me?
Everything you know that has never been written down. The judgment calls, the exceptions, the diagnostic shortcuts, the reasons you deviate from standard advice, the failure patterns you can spot in the first ten minutes of a call. This is extraction work, and it is where the real material is — because the things that feel too obvious to say out loud are usually the things nobody else can say at all.
Ask: what do I do instinctively that I have never had to explain?
Compare Stage Two against Stage One. For every item in the signal inventory, ask whether that knowledge currently exists anywhere in a form that is structured, specific and publicly accessible. Not mentioned in passing. Not implied. Explicitly articulated with enough clarity that a stranger could understand it, apply it, and cite it. Whatever fails that test is the gap, and the gap is the work.
Ask: could someone who has never spoken to me act on this?
The gap made measurable. Six weighted pillars, scored out of 100, plus a twelve-month decay projection that models what remains if you publish nothing further. It turns an uneasy feeling — a lot of this only exists because I am still here — into a specific figure with a specific set of next moves.
Ask: what would remain, and for how long?
The Methodology
One hundred points across six pillars. The weights are not arbitrary — they reflect what retrieval and answer-engine research consistently shows actually determines whether an expert gets quoted, weighted toward the factors that keep working without you.
| Pillar | Weight | What it measures | Why it is weighted this way |
|---|---|---|---|
| Proprietary Signal | 22 | Named frameworks, invented or refined terminology, original data, distinctions that exist nowhere else. | Retrieval favours the highest-concentration source on a topic. Originality is the only durable defence against being one of ten thousand. |
| Definitional Clarity | 20 | Answer-first structure, explicit standalone definitions, tables, named statistics, extractable passages. | Machines quote passages, not pages. Controlled studies put the lift from cited statistics and quotations at roughly 30–40%. |
| Entity Resolution | 16 | One canonical identity: consistent naming, a real author page, Person/Author structured data, sameAs links, visible bylines and dates. |
A system that cannot unambiguously resolve who you are cannot attribute anything to you. Structured data and semantic HTML sit among the pillars most strongly associated with citation. |
| Corroboration | 16 | Third-party coverage, a catalogued book, independent references, your method named by people who are not you. | Earned sources dominate what AI engines cite — one large study put non-paid sources above 95%, with the bulk of that earned media. |
| Coverage Depth | 14 | How much of the real question space in your field your published work explicitly answers — including comparison, cost, and "when not to" questions. | You can only be cited on questions you have answered. Coverage is the surface area of your citability. |
| Custodianship | 12 | A structured archive you own, operable by a successor or a trained system, with a maintenance cadence and a defined custodian. | This is the pillar that converts visibility into permanence. Everything else decays without it. |
Sources informing the weighting include the Princeton/Georgia Tech GEO study (KDD 2024), Muck Rack's analysis of AI citation sourcing, Ahrefs' analysis of most-cited domains, and current answer-engine citation research. The score is a diagnostic instrument, not a guarantee of placement in any specific AI system.
The number matters less than the band it falls in. Each band describes a genuinely different situation, and a different first move.
A single score describes today. It does not describe what happens to today when you stop feeding it — and that is the actual question this framework exists to answer.
So the audit produces a second figure. Each pillar is assigned a persistence coefficient: the proportion of its value that survives twelve months with no new output. A named framework barely erodes. Structured, dated content erodes slowly as freshness signals age. Earned coverage erodes faster as competitors publish. Coverage depth erodes fastest of all, because the question space itself moves. Custodianship does not erode — it is the thing that measures whether anything is being maintained.
The result is your twelve-month projection, and the distance between it and your current score is the honest cost of a body of work that depends on you continuing to show up.
The Legacy Score Finder
Twenty-four questions across the six pillars. Answer as your harshest client would answer on your behalf — the score is only useful if it is uncomfortable. Nothing is sent anywhere unless you choose to request the written report at the end.
Answer all 24 to get a full result.
Your Legacy Score
0/100—
—
Twelve months from today, publishing nothing further
0/100—
Your three highest-leverage moves
Your score is yours to keep either way. If it is useful, we will send the long-form version — every pillar explained, your specific gaps listed, and the order to close them in.
Optional. One email with your report. No list, no sequence, unsubscribe in a click.
The self-assessment gets you the shape of the problem. The full Expertise Audit gets you the extraction, the named framework, and the archive — the part that makes the score move.
Book My ConsultationThe Takeaway
Do not create more noise. Audit the body of work you already have, find the proprietary signal inside it, and make the preservation gap visible. That is the work — and it begins with knowing exactly what you have.
Common questions about the Legacy Score diagnostic and what to expect from the process.
The diagnostic is conducted privately, one to one, with AZMAI. Fill in the form and we will be in touch within one business day to arrange your session.
Availability is limited to a small number of diagnostics each quarter so that every session receives the attention the work demands.
We will be in touch within one business day to confirm your session.
