I Tested My Own AI-SEO Thesis on the Hardest Entity I Could Find: Myself

Two brand-new domains. Almost no traffic. And Google's AI still learned to identify me, describe what I do, and tell me apart from five other people with my name. Here's exactly how — and the part where the same method blew up in my face on a different project.

The setup: I've been making a claim. Time to put it on trial.

If you've read the other two pieces on this site, you know the argument. AI in SEO isn't a tooling problem, it's a retrieval problem. The systems doing the answering — Google's AI Mode, ChatGPT, Perplexity, Gemini — don't rank pages and hand you ten links. They build a model of who you are from signals across the web, then decide whether to cite you. If your entity isn't clear, you don't show up. Not lower down. You're just not in the answer.

That's a confident claim. It deserves a test.

So I ran one on the hardest possible subject: my own name.

"Alex Rodriguez" might be the worst entity-collision name in existence. There's the baseball player. There's a Montevideo-based SEO manager. There's a Houston web designer, an email-marketing director, a UC Berkeley campaign manager. Google's own AI, asked about "Alex Rodriguez SEO," used to surface a lineup of us and tell the user to sort it out themselves.

If entity clarity can resolve *that*, it can resolve anything.

Here's the constraint that makes the result worth your time: I did this on two domains that are days to weeks old, with effectively no traffic. Seven clicks. Eighty-one impressions. My own name sits at position 7.4 with zero clicks on it. That's not a humblebrag in reverse — it's the entire point. Whatever moved the AI, it wasn't authority, backlinks, or volume. There wasn't any. It was structured signal, and nothing else. That isolates the variable in a way a high-traffic site never could.

Before: the AI couldn't pick me out of a lineup

Here's the starting state. Search "alex rodriguez seo," read the AI Overview, and you get a hedge: several professionals share the name, the most prominent being a Montevideo-based SEO manager, and — somewhere in the list — "a professional focusing on workflow, content, and marketing."

That last one is me. Buried at position two of five, described in terms so generic they'd fit half of LinkedIn, with the AI openly punting: *there are multiple people with this name, check the links to figure out which is which.*

That's not a ranking failure. It's an *identity* failure. The machine had no confident model of who I was, so it refused to commit. Every variation of the "produce more content" playbook in the world doesn't fix this, because the problem isn't how much I'd published. It's that nothing told the machine what I *am*.

What I actually did (the part the theory articles skip)

This is the section I'd want if I were reading someone else's case study, so here's the mechanism, not the vibes.

I built a `Person` entity in structured data on the named domain and wired it into an organization graph. Concretely, three layers — the same three layers I argued for in the framework piece, applied to myself:

1. Entity definition. A JSON-LD `Person` node: full legal name with given/family split, alternate names ("AlexSEO," "Alex Rodriguez SEO"), job title ("AI-First SEO Operator"), location in Cedar Park, TX, an image, a real description of the actual specialization. This is the disambiguation engine. It hands the machine the precise fields it needs to separate me from the baseball player and the four other SEOs.

2. Relationship mapping. The entity doesn't float. It's wired into an organization graph via `worksFor`, `founder`, and `parentOrganization` — connecting me to MFGSEO LLC, recording the founding date, the corporate lineage, and the related learning community. This is what lets an AI state not just who I am but how the businesses connect.

3. Topical authority. A `knowsAbout` array declaring the actual expertise space — Answer Engine Optimization, entity-based SEO, generative engine optimization, schema markup, AI Overview optimization, knowledge graph optimization, content architecture — plus a service catalog. This tells the machine what I'm an authority *on*, in machine-readable terms, instead of hoping it infers it.

The connective tissue across all of it is a `sameAs` array: the agency site, the public learning community, LinkedIn, a dedicated Reddit community I created, YouTube. Consistent name, bio, and specialization repeated across every surface AI engines pull from — so the machine can collapse all those profiles into one confident entity instead of guessing they might be different people.

Then I validated it was actually live and readable. Google's Rich Results Test confirmed the domain crawled cleanly with valid Organization and LocalBusiness structured data. The schema isn't theoretical. Google has it.

No new backlink campaign. No content volume push. No PR. Just a clean, consistent, machine-readable entity, declared once and corroborated across the surfaces that matter.

After: the machine resolved me — and disambiguated me on its own

Same query. "Alex Rodriguez SEO." Different universe.

The AI Overview now leads with me, by full name and handle, and describes me specifically: an SEO professional specializing in AI-driven content strategy and AI-first SEO, with over 15 years of experience, focused on building content systems designed to get cited as sources in AI-generated answers — moving beyond traditional ranking metrics.

Read that description again. That's not generic. That's my actual thesis, repeated back to me by the machine, because I put it in the schema.

Then it does the thing that proves the entity truly resolved. It disambiguates *on my behalf*: this is Alexander Rodriguez the SEO specialist — not to be confused with the others of the same name in sales or digital campaigns.

That single sentence is the whole case study. The machine went from "there are several of you, you figure it out" to "here is the specific one, and let me separate him from the rest for you." That is entity clarity being won, on camera, on a domain with no traffic.

And it's consistent across queries and surfaces:

• Ask about me directly and the AI Overview now states the positioning verbatim — a recognized SEO and AI systems strategist with 15+ years of experience — with my own content ranking right beneath it.
• Ask who owns MFGSEO and Google traces the corporate graph correctly: founded and owned by me, established as an offshoot of the original parent agency, based in the Cedar Park / Austin area, focused on diagnostic, sprint-based work. It knows the *relationships* — because I declared them.

Each of those maps to a layer I built. Entity definition → it names and describes me. Relationship mapping → it traces the company lineage. Topical authority → it states my expertise and ranks my content for it.

The honest part: what this proves, and what it doesn't

I'm going to be precise here, because the entire reason this project exists is that most "AI cited me!" posts overreach and fall apart under one glance at the data.

What it proves: structured entity signals alone — on near-zero-traffic, days-old domains — got Google's AI to correctly identify me, describe my actual specialization, trace my business relationships, and actively disambiguate me from five other people with my name. Entity clarity moved *independently of, and prior to,* traffic and rankings. That's the finding.

What it does not prove: that this drove pipeline, traffic, or revenue. It hasn't. Not yet. The clicks aren't there — the domains are too young and the search volume on the queries is tiny. Anyone who pulls my Search Console will see exactly that, and they should, because it's the proof that the entity result wasn't bought with volume.

If I told you the leads poured in, I'd be doing the precise thing I built this practice to call out: a confident claim with nothing under it. The discipline is the credibility. Entity resolution is the leading indicator. Traffic and pipeline are the lagging ones, and I'll show those when they're real — not before.

The counterweight: the same method, applied wrong, cratered a different site

Here's where I show you the framework's failure mode, on my own asset, because a case study that only contains the wins isn't a case study. It's an ad.

I own a programmatic SEO system — geo-targeted city pages for a single commercial service, built from one architecture across more than twenty Texas territories. Same thesis: templated delivery, real entity signals, scaled across geography. And for a while it worked. Several city pages ranked in the single digits. The core commercial query sat at position 3.

Then I broke it. Two ways.

I migrated platforms and snapped the continuity. Clicks went from 107 in one month to 3 the next. Average position drifted from the low 30s into the high 50s. Migrations are the highest-risk thing you can do to an indexed site, and I hit the classic failure modes — dual URL schemes live at once, prerender continuity gaps, redirect chains that didn't carry the old signals to the new URLs. The machine had to re-evaluate from scratch, and it didn't like what it found.

I diluted the entity into mush. This is the more instructive failure. The site started as a focused, single-service geo-system — clear entity, clear expertise, clear answer to "what is this site?" Then I expanded it into dozens of unrelated services: roofing, mosquito control, short-term-rental cleaning, epoxy floors, garage doors, generators. The query data turned into noise. The original commercial intent got buried under positions in the 40s through 90s. The machine could no longer tell what the site *was* — and an entity it can't define is an entity it won't cite.

That's the boundary condition of the entire thesis, learned by violating it: entity clarity scales with focus and dies with dilution. Templating works when you template *depth* — one thing, done with real signal, across many places. It collapses when you template *breadth* — many unrelated things sharing a domain, each diluting the others until the entity is unreadable.

I proved both directions in the same month. The focused entity (my name) resolved with almost no traffic. The diluted entity (the over-expanded site) lost 97% of its clicks despite having a year of history. Same operator. Same framework. Opposite results, for reasons the framework predicts.

The refined thesis

Visibility is organization.

Not volume. Not effort. Organization — in the literal, structural sense. Whether the machine can build a clean, confident, consistent model of who you are, what you do, and how you connect to everything else.

When the signal is focused and consistent, the AI resolves you fast, even from a standing start with no traffic. When the signal is sprawling or contradictory, no amount of history or content saves you — the machine loses the plot and routes the citation to someone clearer.

The test that matters isn't your rank-tracker. It's whether an AI engine, asked about you, can say who you are with confidence and tell you apart from the noise. I now have screenshots of that going both ways — the win and the failure — run by the same person, the same month, on real assets.

That's the proof the framework needed. Here it is.

Want to know whether AI engines can confidently identify your business — and whether they're routing your citations to a competitor with a clearer entity?

That's the first question in every diagnostic I run