Originally explored on Jax5d, December 2025 — adapted and expanded for Probably Genius.
AI can’t cite what you won’t claim. AI assistants are attribution machines: they build answers from sources they can name, verify, and cross-reference, and expertise that was never claimed under a real name is invisible to them. Claiming is documentation, not self-promotion: named positions, a published method, and one consistent identity on every surface the machines read.
If you are the expert who does the work but never takes the byline, this will land uncomfortably. No published stance, no named methodology, no counted wins, because specifics feel like showing off and the work should speak for itself. Then one day you ask an AI assistant about your own specialty and watch it hand your category to someone louder. Not better. Louder, and claimed.
The filter is unsentimental, and it is measurable. Semrush studied 11,882 prompts across ChatGPT Search, Google AI Mode and Perplexity and found strong E-E-A-T signals carried a +30.64 percent correlation with getting cited, among the strongest factors it measured. The machine quotes whoever put their name on the answer. Expertise nobody claimed reads as nobody’s.
Our co-founder Jacquie Baker first published this argument in December 2025, on her personal Substack, after months of watching the same wound turn up in almost every engagement. The gaps were never missing information. The clients had the credentials, the track records, the outcomes. What they did not have was the willingness to write any of it down under their own name.
This is the expanded version of that argument for business owners: why attribution needs a claimant, what claiming looks like in practice, what modesty costs, and how to claim your expertise without becoming the person at the barbecue nobody wants to stand next to.
Attribution Requires a Claimant
Most people hear “AI search optimization” and think keywords. Gaming algorithms. Stuffing metadata like it’s 2012. That mindset belongs to a machine that no longer decides anything. The systems answering your buyers’ questions work on a different principle: AI doesn’t rank pages. It understands entities. It builds a model of who you are, what you’ve done, what you believe, and who you’ve helped, and then it decides whether that model is solid enough to stand behind out loud.
Which means the question was never “how do I trick the algorithm?” The question is: how do I help AI understand who I actually am? And that question has a prerequisite most marketing conveniently skips. You have to know first. You have to be willing to say it.
Millions of people spent the first years of consumer AI learning to treat an assistant as a late-night confidant, the thing that listens at 3am and remembers the whole backstory. That trust is the foundation of the next buying behavior, because the same assistant is becoming the advisor that recommends which accountant to hire, which architect to call, and which sleep consultant might actually fix your insomnia.
You have a dispute you have been processing with your AI for months; when you finally need a lawyer, you do not start from scratch. You get three names in your price range with reasons attached. No search results page. No asking friends. A single recommendation from a system the buyer already trusts.
This is where the AI wars will be won. Not in the boardrooms. In the quiet moment when someone trusts their AI’s recommendation more than a search result.
And it can only recommend what it understands. When Ahrefs analyzed 1.4 million ChatGPT prompts, it found the model went on to cite only 49.98 percent of the URLs it retrieved, with titles, snippets and source identity doing the heavy lifting in the first cut. Half the candidates never get fully read.
What survives is the material a cautious system can attribute: a named person, a stated method, a specific outcome, a story that checks out everywhere it looks. That is the attribution economy in one sentence. The engines quote whoever put their name on the answer, and a category whose real expert stayed quiet gets credited to whoever claimed it first.
This is the layer of the market we measure as AI visibility: not whether your website ranks, but whether the machines mention, describe, and recommend you when a buyer asks. The rest of this essay is about the single biggest reason good businesses fail that test.
What Claiming Actually Means
Google’s framework for evaluating content is called E-E-A-T: Experience, Expertise, Authoritativeness, Trustworthiness. Strip the jargon and it is four questions. What have you actually done? How long have you been doing it? What specific outcomes have you created? Why should anyone believe you? Simple questions. Most people choke on them.
Not because they don’t have answers. Because claiming those answers feels like too much. Writing your bio feels like a job interview with yourself. Counting your wins feels like bragging. Being specific about your outcomes means being seen. So people write “passionate about helping others” and call it a day, and the machines respond to that sentence exactly the way it deserves: they find nothing in it to verify, nothing to attribute, nothing to quote.
So here is claiming, made practical. In our work it is three moves.
Named positions. A point of view published under your own name. What do you believe about your field that some of your peers would argue with? Where is it written down? A byline is a claim; an anonymous services page is not.
Published methods. The thing you do in the first ten minutes of every client call is a methodology. Until it has a name and a page, it does not exist to a machine.
One consistent identity. The same name, the same story, the same specialty on every surface the engines consult, so the machine’s file on you gets one coherent answer instead of four fragments. That identity layer is its own discipline, which is why we publish our method for it as Entity Identity.
The data backs the specificity part bluntly. In the Semrush study, the single strongest positive correlation with getting cited was clear summarization at +32.83 percent, and the peer-reviewed GEO study presented at KDD 2024 found that adding citations, quotations and statistics lifted a source’s visibility in generative engine answers by up to 40 percent, while old-style keyword stuffing made it worse. A dated number under a named author is machine-food. An adjective is air.
Notice what all three moves have in common: none of them requires you to become someone else. They require you to document who you already are.
The experts with the deepest experience are usually the worst documented, because they were busy doing the work, and their genius lives in war stories, judgment calls and phone-call explanations no machine can read. We named that pattern the Black Box Effect, and claiming is the way out of the box: moving what is true about you from your head onto the record.
What Modesty Actually Costs
Here is the price list. If you won’t claim it, AI can’t either. If you’re vague about your value, AI will be unsure how to recommend you. And if you hide your wins because you don’t want to seem full of yourself, you disappear into the sea of AI slop, the infinite content generated by people who figured out the tools but have nothing real to say.
The engines are drowning in that material, which is precisely why they cling to what they can verify and attribute. Modesty does not read as confidence to a machine running an identity check. It reads as absence.
The gaps in your EEAT aren’t information problems. They’re claiming problems. That distinction matters because information problems get solved with more content, and claiming problems get worse with more content, because every unclaimed article deepens the pile of material the machines cannot pin to anyone.
The original essay told a story worth carrying forward, with the client’s details kept private. He had folded his previous business when the rules of his industry changed. The guilt was heavy. He had gone from national growth lists to hiding, and the story he told himself was the one shame writes: I failed, I’m starting over, don’t look too closely.
Then the work started with counting. Not spin, arithmetic: the engagements he had personally delivered, the clients served, the value moved, over years. Laid out in a row, it was not a failure. It was a track record interrupted by a rule change. His positioning was rebuilt around what was verifiably true, and the machines started citing him cleanly, because there was finally something claimed to cite.
But the part that stays with you is not the citations. Armed with the counted record, he went back to old business partners, bracing for scorched earth, and found none. They understood what had changed in the industry. The partnerships reignited. The shame he had been carrying for years was gone in a conversation he had been too afraid to have.
That is what the numbers are actually for. Yes, AI needs specifics to cite you confidently. But the deeper shift happens in the person who finally sees their own evidence assembled in one place. The old story gets replaced by the record. Oh. Right. I’ve been doing this. Successfully. For years.
This is why we say the client is the genius and we are the translators. Nothing in that engagement was invented, inflated, or optimized into existence. It was counted, claimed, and put where the machines could read it. The expertise was always real. The claim was the only thing missing.
How to Claim Without Bragging
The fear underneath all of this deserves a straight answer. Bragging is performance: adjectives about yourself, arranged for applause. Claiming is documentation: facts about your work, arranged for verification. “Award-winning visionary” is bragging, and the machines ignore it because there is nothing to check. “I have prepared tax filings for construction firms since 2011, and I published my method for job-cost cleanups” is claiming, and a machine can verify every word of it. The paradox of the attribution economy is that the humble-sounding sentence is the specific one.
The practice, in four steps.
One: count what is true. Sit down with your calendar, your invoices, your files, and inventory the years, the engagements, the outcomes you can stand behind under questioning.
Two: name your method. You already have one; you perform it every week. Write it down, title it, publish it on a page you own.
Three: put your name on positions. Publish the professional opinions you defend in rooms, under a byline, dated.
Four: say it the same way everywhere. One name, one description, one story, on your site, your profiles, your directories, so the machine’s file on you converges instead of splintering.
Then hold the discipline that makes claiming safe: never claim past the evidence. A claimed exaggeration is worse than an unclaimed truth, because attribution cuts both ways and the machines keep receipts.
We hold our own published work to a version of this standard, the AI Integrity Standard, a 100-point assessment where anything below 85 does not publish. You do not need our scale to adopt the principle. Every claim you publish should survive a skeptical stranger checking it, because that skeptical stranger now exists at scale, answers questions all day, and decides whether to say your name.
Which brings the whole argument down to the question underneath the question. Everyone is going to be asking how to show up in AI search. But the real question is older and harder: Do you believe your own story? Because if you don’t, no algorithm can save you.
And if you do, if you will actually claim it, count it, structure it, and put it on the record, you become uncopyable. Not because of SEO tricks. Because you remembered who you are.
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. Jacquie has been publishing the thinking behind this work on her Substack, Jax5d, since 2025; this essay adapts one of those arguments for the businesses it was always really about.