THE RECOMMENDATION LAYER · AUGUST 6, 2026 · 9 MIN READ

Why Doesn't My Best Expertise Show Up When Buyers Ask AI?

Your best expertise is invisible to AI because it was never written down. Spoken judgment leaves no trail a machine can verify, and unpublished skill cannot be cited.

By Jacquie Baker
TL;DR: WHAT TO REMEMBER

Your best expertise never shows up when buyers ask AI who to trust because it has never been written down where a machine can check it: the judgment that wins your work lives in conversations, and conversations leave no record to verify. The engines are not ranking your skill. They are assembling evidence, and in findings published in April 2026, Ahrefs found ChatGPT cites only about half of the pages it even retrieves. Material that was never published is not late in that queue. It was never in the building.

You already said the thing that would win the next client. You said it on Tuesday, on the difficult call, when the numbers did not add up and you caught the risk before anyone else did. The client nodded. Nobody wrote it down. Next month a stranger asks an assistant who to hire for exactly that problem, and your name is not in the answer.

Not because you are worse. Because the machine never got to hear you. You weren't rejected. You were omitted, and omission is a filing outcome, not a judgment of the work.

This piece walks through why spoken expertise cannot reach an AI answer, the three checks engines run before they name anyone, and how specialists get their judgment onto the record without becoming writers.

The Tuesday Problem

Every genuine specialist has a Tuesday version of themselves: the one who shows up in the working sessions, on the walkthroughs, in the calls where something is going wrong. That version draws distinctions in seconds, catches the risk nobody briefed, explains a hard tradeoff so plainly the client repeats it at dinner. It is the reason your clients stay for decades.

Ask that specialist to describe their website and a different person appears: experienced, dedicated, client-focused. The Tuesday brilliance is nowhere on it. When we ask why, the answer is almost always the same, and it is almost touching: it is obvious. Why would anyone publish the obvious?

Obvious is doing a lot of hiding in that sentence. The distinction you consider table stakes took you twenty years to learn, does not appear on any competitor's site, and is precisely what a buyer, or a machine assembling an answer for one, would treasure. Inside twenty years of expertise, everything looks obvious. From the outside, that same material looks like the answer.

Referrals hide the cost of this for years, which is why the problem surfaces so late. As long as introductions arrive through people who were in the room, the room is the record, and it works. But the buyers coming through assistants were never in the room, and the referral partners who would have vouched for you are not part of the answer the machine assembles. The better your word-of-mouth economy has been, the less written evidence it ever needed to leave behind.

So the calendar fills by referral, the work stays excellent, and the public record says almost nothing that a hundred other firms' records do not also say. The best of you airs live, once, to an audience with no notebook. The record never hears it.

Respected in the Room, Invisible to the Record

Reputation, as you have built it, is a human network effect: colleagues who have seen your work, clients who tell stories, referral partners who vouch by phone. Rich, earned, and almost entirely oral. An AI assistant cannot subpoena any of it. When a buyer asks who to trust, the engine works from what is published, structured and corroborated, which is to say, from the thin version of you.

This is not the machine being stupid. It is the machine being careful. Recommending a stranger to a stranger is a risk, so the engines favor what they can verify over what anyone merely asserts, the audition we walk through in how ChatGPT decides who to recommend. An unwritten reputation, however real, is unverifiable by definition.

Here is the diagnosis in one line: the gaps in your E-E-A-T aren't information problems, they're claiming problems. The information exists in you. It has simply never been claimed, on the record, under your name, in a form that can be quoted. AI can't cite what you won't claim.

And claiming has a compounding rival: everyone else's photocopies. The internet already holds the generic version of your category's advice, restated thousands of times. You can't photocopy a photocopy and expect it to be worth citing. What the machine lacks is not more restatement. It is the high-resolution photograph only your practice could have taken.

The Three Checks You Silently Fail

When an engine considers carrying your name into an answer, three checks run, and spoken-only expertise fails all three quietly.

Notice the pattern: each check fails for filing reasons, not merit reasons. The expertise exists, the goodwill exists, the distinctions exist. None of them are on the record, so, on the checklist the machine runs, none of them exist at all.

Corroboration trips even the well-published, which shows how strict the checks run. In the June 2026 Lily Ray analysis of 100 B2B software queries, Google's AI Overviews cited a brand's own listicle and still recommended someone else in 69 percent of the cases studied. A page can pass the reading test and fail the vouching test in the same answer.

Run the test on yourself in one minute. Take the single distinction you explained most recently to a client, the one that changed their decision, and search for it under your name. If the answer only lives in that client's memory, you have found the exact spot where the recommendation went to somebody else last week.

Google says the quiet part in its own documentation, asking whether content demonstrates first-hand expertise and a depth of knowledge, and warning against pages that mainly summarize what others have said. The systems are actively hunting for exactly what you have. They can only reward the version of you that got written down.

What Translation Actually Looks Like

The fix is not becoming a content creator, and the machine does not need you to perform. We do not replace the expert with content. We turn expertise into evidence, and the raw material is conversation, the thing you already do all day.

The Genius Interview is that conversation, recorded: your hard cases, the calls you make differently, the risks you catch early. From one interview a month come 12 done-for-you thought-leadership articles, planned to produce approximately 60 clear Knowledge Entries: your distinctions, published as direct answers to the questions your buyers actually ask, structured so a human or a machine can lift them whole and attribute them to you. The mechanics of the pre-call payoff are in how thought leadership builds trust before the first call.

The quality bar matters double here, because a specialist's record must never say something the specialist would not. Every claim traces to your approved facts or a real source, and every piece is scored 0 to 100 through the Integrity Gate: nothing publishes under 80. The test we apply to every draft comes from our own writing canon: if ChatGPT could have written the piece without talking to you, it is a photocopy, and it does not ship.

What accumulates is the thing your reputation never had: a public, verifiable, growing body of your actual judgment. The only content AI has a reason to quote is knowledge it could not have generated alone, and your Tuesday self produces that in every consult. Translation just stops letting it evaporate.

The Record Starts With One Conversation

Here is what changes when the record catches up with the room. Buyers arrive having already met your judgment, so the first call starts at fit instead of proof. Referral partners get a page to point at instead of a phone number to vouch for. And the engines, asked who to trust with exactly the problem you are best at, finally have something under your name to check.

The compounding matters as much as the first month. Every published distinction becomes permanent, findable and quotable, which means the record only moves in one direction while the spoken version keeps evaporating on schedule.

Whether they then say your name is not something anyone can promise, and we do not. The work makes you verifiable, corroborated and worth quoting, then measures what happens: the same buyer questions, five engines, month over month. Readiness is the input. The answers are earned.

Start by seeing what the machines currently know. The free Recommendation Check scores your visibility across 109 checkpoints and records what five AI engines answer when buyers ask who to trust in your category. For most specialists it is the first time they have ever seen the thin version of themselves the machine has been working from, and it explains a lot of quiet phones in a single page.

The expertise was never the problem. The filing was. That is fixable, starting with one recorded conversation about the work you already know cold.

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. AI is the amplifier. The clients are the signal, and the machine is never the most interesting character. You're probably a genius at what you do. We make sure AI gets the memo.

Why doesn't my expertise show up in ChatGPT or Gemini answers?
Because AI engines can only work from what is published, structured and verifiable, and most deep expertise never gets written down. The judgment that wins your work happens live, in consults and difficult calls, and leaves no public trace. When a buyer asks who to trust, the engine assembles an answer from the written record, so it names whoever left the clearest trail, not necessarily whoever knows the most. The gap is a filing problem: real expertise with no published, checkable version of itself is invisible to the machine by definition.
What does AI check before recommending a business?
Three things, in essence. Existence: that the business is one specific, confirmable entity, with a name, people, credentials and locations that agree everywhere the machine looks. Corroboration: that independent sources, not just the business itself, tell the same story in writing. And information gain: that the business's published material knows something the rest of the internet does not already say. Spoken-only expertise fails all three quietly, because none of the evidence exists where a machine can read it.
Do I have to become a content creator to fix this?
No. The fix is translation, not performance. Your raw material is conversation, which you already produce all day; it just needs capturing and publishing in a verifiable form. An operated program records a structured interview about your cases and judgment, then produces 12 done-for-you thought-leadership articles a month, planned to create approximately 60 clear Knowledge Entries under your name, each quality-gated before it ships. Your time cost is about an hour a month. No camera, no calendar, no personal brand required.
How do I find out what AI engines currently say about my business?
Ask them the questions your buyers ask, across several engines, and record the results, because answers vary by engine and by week. The free Recommendation Check does this systematically: it scores your visibility across 109 checkpoints and captures what five AI engines, ChatGPT, Gemini, Perplexity, Claude and Grok, answer when buyers ask who to trust in your category. The report shows what the machines can currently verify about you and where your real expertise never made it onto the record, before you spend anything fixing it.

CITATIONS

  1. "Why ChatGPT Cites One Page Over Another (Study of 1.4M Prompts)" (Ahrefs, published April 2026). Analysis of 1.4 million ChatGPT prompts: the model cites 49.98 percent of the URLs it retrieves, with titles, snippets and URLs filtering candidates before page content is read. Evidence that AI answers are assembled from a competitive, verifiable written record, which unpublished expertise never enters. ahrefs.com
  2. "Creating Helpful, Reliable, People-First Content" (Google Search Central). Google's guidance asks whether content demonstrates first-hand expertise and a depth of knowledge, and warns against content that mainly summarizes what others have to say. The official confirmation that the systems are looking for exactly the practice-earned judgment most specialists never publish. developers.google.com
  3. "Google AI Overviews cite self-serving listicles, but recommend competitors 69% of the time" (Search Engine Land, June 2026). Lily Ray's analysis of 100 B2B software queries found 323 cases where an AI Overview cited a brand's own listicle and, in 69 percent of them, recommended someone else. Even published content fails the recommendation checks when the surrounding record cannot corroborate the brand behind it. searchengineland.com
WRITTEN BYJacquie ("Jax") Baker

Founder of Probably Genius, an AI visibility firm helping professional service brands become the named answer in AI search. Nearly two decades across technology, digital strategy and branding, including 1,000+ digital projects through her previous agency, now focused on making experts visible, verifiable and recommendable to AI. Let's talk →

Related reading

Read the method.
Then see it run on you.

The 109-point diagnostic is this library, applied to your business. Free, about an hour to present, no obligation.

See where you stand →

A real diagnostic, not a sales call in disguise.